COURSE I LABORATORY

GUIDE TO
REPORTING AND MEASUREMENT
(version: 2018)

Please keep this guide safe and in good condition as you will use it
throughout your 1st and 2nd semester laboratory activities.

Name: ____________________________
 Guide to Reporting and Measurement: January 2018

CONTENTS
FOREWORD .............................................................................................................................................................3

SECTION A – PERFORMING AN EXPERIMENT .........................................................................................................7

A1 Planning an experiment ................................................................................................................................7

A2 Writing a laboratory report...........................................................................................................................8
A3 Drawing tables and graphs ........................................................................................................................ 11
A4 Significant figures ....................................................................................................................................... 14
SECTION B – MEASUREMENT UNCERTAINTY ...................................................................................................... 15

B1 Overview .................................................................................................................................................... 15
B2 TYPE A evaluation of uncertainty ............................................................................................................... 17
B3 TYPE B evaluation of uncertainty ............................................................................................................... 19
B4 Combination of uncertainty in one measurement .................................................................................... 22
SECTION C – CALCULATIONS WITH UNCERTAINTIES ........................................................................................... 26

C1 Calculation of uncertainty across multiple measurements ....................................................................... 26

C2 Examples of uncertainty calculations......................................................................................................... 27
SECTION D – REPORTING RESULTS ...................................................................................................................... 30

D1 Quoting a result ......................................................................................................................................... 30

D2 Comparing results ...................................................................................................................................... 33
D3 The uncertainty budget ............................................................................................................................. 34
SECTION E – FITTING A LINEAR FUNCTION TO THE DATA ................................................................................... 36

E1 The process of fitting a function to a set of data ....................................................................................... 36

E2 Linearising equations ................................................................................................................................. 37
E3 Method of least squares ............................................................................................................................ 38
E4 Using EXCEL (version used in PHYLAB1) ..................................................................................................... 39
E5 Using LINEARFIT ......................................................................................................................................... 40
SECTION F – EXAMPLE OF FITTING A GAUSSIAN ................................................................................................. 41

Back cover: Oscilloscope instructions for Practicals – Part II.

2
FOREWORD

Welcome to the Physics Course I Laboratory

As a science or engineering graduate, it will be assumed by your future employer that you have certain skills.
These will include:
 problem solving skills;
 the ability to engage with apparatus that you may not have seen before;
 the ability to plan and execute an experiment or some type of investigation;
 the ability to collect, analyse and interpret data, and
 the ability to communicate and present your findings either orally or in the form of a written
report.

Developing skills in scientific measurement

Physics is about modelling and understanding the phenomena of nature, and physics experiments are
concerned with the creation of new knowledge through measurement. These measurements are used to
formulate new physics theories, or test existing theories. However, there are many factors that influence the
result of an experiment, for example factors associated with the calibration of the apparatus you are using,
how you used the apparatus, environmental effects, etc. Each of these factors work against you getting a
“perfect” result in your experiment.
During this course you will learn how to report a scientific measurement result. In the same way that there
are mathematical tools of physics such as coordinate systems, vectors, force diagrams, etc. which are
universally accepted ways of communicating physics models, there are clear guidelines regarding the way in
which scientific measurement results should be reported. These have been agreed upon by all the
international physics, chemistry and other science organizations, and are methods you will be taught in this
course. The way that you learn to analyze your experiments in this course is the same way in which scientists
all around the world do so. For example, you will learn the scientific methods of numerically estimating
uncertainties in an experiment in a realistic way. “Uncertainty” in a science measurement is a parameter (a
number) that comes from analyzing all the factors that you think have influenced your experiment. This
parameter, together with the best approximation of the measured, form part of the measurement result. You
will see in this course that the measurement result is a compact way of summarizing all the knowledge that
you have about a measurand. A measurement result is a statement of probabilities. We can never know the
true value of a measurand: we can only make statements about our knowledge about the interval in which
the measurand exists.
Furthermore, since the nature of the enterprise of physics is about the creation of new knowledge, the more
“careful” you are in a physics experiment, the smaller your uncertainty should be. Therefore, scientists design
and carry out experiments so as to have as small an uncertainty as is realistically possible. The smaller your
uncertainty is, the better knowledge you have about a particular measurand. Of course, it is impossible to
make a measurement which has zero uncertainty.
Having an interval as the result of an experiment, and not a single point, also allows us to compare two
measurement results with each other, or with a theoretical (calculated) value. For example, we can say that
the two measurement results 3.4 ± 0.6 J and 3.7 ± 0.4 J agree with each other, while the measurement
10.41 ± 0.07 m s-2 does not agree with the accepted value for Cape Town of 9.79 ± 0.01 m s-2.

Safe laboratory practice

The Physics Course I Laboratory is a safe teaching laboratory and we expect you to approach all activities in
the laboratory with a “safety first” attitude – this means that the first thing you consider upon entering the
laboratory is the safety of yourself, the other people in your class, the staff and equipment, and also safety is
a priority and important to you.

3
 Guide to Reporting and Measurement: January 2018

You can take these safety skills to whichever scientific field you may venture. Here are three basic rules of the
laboratory:
 You must wear enclosed and sturdy shoes, and bare feet and/or open shoes (e.g. flip-flops) are not
allowed in the laboratory. This rule minimises the risk of you injuring your feet, such as due to broken
glass or equipment dropped from the benches.
 No eating is allowed in the laboratory and drinking is only allowed from closed containers. It is safer
to take a short break from the laboratory if you feel the need for refreshment. Occasionally hazardous
substances such as radioactive sources are used in the laboratory, therefore this rule minimises the
risk of you injuring yourself due to an exposure to a hazardous substance. It also minimises damage to
equipment, such as the damage of electrical equipment due to water contact, and maintains the
laboratory as a clean space to work in.
 Behave professionally in the laboratory and be considerate and respectful of students in nearby
groups. More workplace accidents happen when people fool around than when they carry out their
normal work.
A useful reference to read more about laboratory practice is Etiquette in the Laboratory (1976), American
Journal of Physics, 7 (44). See http://scitation.aip.org/content/aapt/journal/ajp/44/7/10.1119/1.10320

Laboratory staff
During practical sessions, the laboratory space is managed by the following people:
 The academic in charge;
 The chief scientific officer and scientific officer;
 The lab manager, and
 The demonstrators.
You can ask any of the people above for help during the course of a practical, for issues such as conducting an
experiment, administration of the laboratory course, concerns about safety and faulty or damaged
equipment. If you are absent from a practical or tutorial or want to query a tutorial or practical mark
entered on WebApp or Vula, get in touch with the lab manager; Mr. Mark Christians.

Pre-practical preparation
Before you attend each practical session, read through the material on the practical that you are about to
perform and take a look at the relevant instructional videos in the link given below. You will use the practical
time the most efficiently if you come prepared and ready to ask questions.
http://www.phy.uct.ac.za/phy/courses/PHYLAB1.

Practical groups
It is intended this course is undertaken in cooperative learning environment. We believe that working in
groups of 3 is best for these activities. Compare your responses to those of your work partners. Help each
other to figure out what is going on. Resolve any difficulties that you might have and call a demonstrator if
you need to.
You can choose your own group at the beginning of the semester. We use these groups to allocate and
schedule the marking of laboratory reports, so please submit a registration form for your group to the lab
manager in the first session of the semester. This will remain your group, and your laboratory day, for the
rest of the semester. However, if you want to change your group at any time and for any reason, please talk
to the lab manager who can assist you in arranging a different group and change in registration, as well as
making sure your marked lab reports are returned to the correct session.

4
Some guidance on how to work in groups is given in the following pages:

What I really want to know is

why we have to work in Well, companies and institutions that employ science
groups of three during our and engineering graduates rate the ability to work as
part of a team as one of the most important skills
physics practicals! they want their employees to have. Knowledge and
technical skills are of no use if you cannot apply them
in cooperative interactions with other people.
I am grateful that when I was in first year last year I
was given the opportunity to work with other
students in class. I am very shy to ask questions
during lectures, but when I am working in a group I
am able to ask all the questions I need to.

I am sure that you will agree with

me when I say that I understand
something better when I am
Learning to work with others is actively involved in a learning
fine, but what about my activity. When I just sit in class
physics? I am determined to do and listen to the lecturer I cannot
well. remember much of what is said,
no matter how interesting the
lecture is.

I also thought so at first. However,

after a while I found that as a group
we were able so solve far more
difficult problems than I could do on
my own. We were also able to
practice the problem solving
But won’t my marks be strategies that we need to master so
poorer if I have to work with that when I was alone I found that I
my class mates? didn’t get stuck as often.

It’s like learning to play a new

game! No matter how many times
Mmmm..... you watch and listen to the coach,
you only learn when you have a go
I see what you mean.
with a friend. It is just the same
with your physics.

5
 Guide to Reporting and Measurement: January 2018

You know, sometimes Don’t worry. Resolving that type of

I don’t want to explain confusion is an important part of the
something to someone else learning process. When you talk
because I am afraid that I about the physics in a group you will
might confuse myself. come away with a better
understanding of both the abstract
concepts and the problem solving
techniques.

What if I am put in
a group with people
that I don’t get on I have found that even when I think
with too well? that I don’t get on with someone,
after a while when we have
focused on what we have to do,
things actually turn out fine. I have
rarely met anyone that I can
absolutely not work with.
But anyway, the groups don’t stay
together permanently, so you
won’t be with the same people
forever!

What can I do
so that our group works
well together?

There are a number of things that you can do in a

group that will help you to work together better.
The most important is that everyone has a turn to
participate and explain their understanding and
reasoning. You therefore need to contribute and
listen carefully to what your group members are
saying.
Check that everyone understands before moving
on. Don’t exclude people by using a language that
one person doesn’t understand.

When you are stuck on a question, there are a number of questions that you
should ask yourselves before you call a demonstrator:
 Do we understand the question?
 Have we consulted the Guide to Reporting and Measurement?
 Would it be useful to draw a diagram?
 What question are we going to ask the demonstrator?
... this is very important ... it’s no use saying to the demonstrator,
“We’re stuck” ... or ... “We don’t know what to do.”

6
SECTION A – PERFORMING AN EXPERIMENT

A1 Planning an experiment
There are a number of stages to an experiment, each requiring you to think about different things.
Stage 1. Identify the problem that needs solving, or the question that needs answering. What is the
aim of the experiment? Think about what results you need in order to make a valid
conclusion.
Stage 2. Plan your experiment carefully (see notes below).
Stage 3. Do the experiment and complete the analysis of the data.
Stage 4. Formulate your conclusions to the experiment based your results.
Stage 5. Communicate your experiment in the form of a report.
Never rush into taking readings without careful planning. Of course, it is a good thing to familiarise yourself
with the apparatus before starting, but before doing serious measurements sit down and plan carefully
exactly what you are going to do. In most cases you will be working with one or two other students in the
laboratory. It is important that you discuss things with them. Listen to each other’s opinions and talk to a
demonstrator, if necessary.
Some important considerations are:
 What are the main steps that I need to carry out? A flow chart is often useful.
 What apparatus do I need and how does it work?
 Are there any precautions that I need to take?
 How much time do I have for the experiment?
 What variables are involved and what exactly must I measure?
 How many measurements do I need to make?
 What tables must I draw up? Detailed tables should be drawn up before taking readings as the tables
will serve as a guide.
 What are the main steps in the analysis? Again, a flow chart is often useful.
 Which graphs (if any) will I need to plot?
 What influences will affect the measurements? Make a list beforehand and add to it as you proceed.
After the experiment, reflect on how well your plan worked, noting both successful aspects and failures of the
plan. Keep in mind the structure of your report that you will have to write. Remember that if, while writing
the report, you suddenly remember that you did not record something important while doing the experiment,
it will be too late!

7
 Guide to Reporting and Measurement: January 2018

A2 Writing a laboratory report

The purpose of a laboratory report is to communicate the aim, process and outcome of an investigation to an
outside audience. It is a record of your direct (“hands-on”) experience in the laboratory. In most cases, a
scientific investigation is considered to be incomplete without a report. By synthesising (putting together) the
different aspects of your laboratory experience in a structured and coherent report, the essence of your
investigation becomes clearer in your own mind. In the process, you develop your skills of reasoning and
ability to communicate in writing.
There are many acceptable ways of presenting a scientific laboratory report, but, almost all reports will
include the components outlined here.

Components of a laboratory report

Title:
This will include the author’s name (your name), your partner’s names, the course code, the date of the
experiment and a suitable title of the report.
The title must be short but factual and descriptive. It must summarise the major aspects to be dealt with in
the report. The key words will often come from the laboratory task that has been set and you need to identify
these. These words help clarify the requirements of the task and also alert the reader as to what the report is
about.
Introduction and aim:
The introduction puts the report into perspective by giving the reader relevant background information about
the phenomenon being investigated. This is in order to prepare the reader for what he or she is about to read.
This background may include some historical information or developments, earlier experimental work, the
theory or law governing the phenomenon being investigated. These introductory remarks must be kept brief
to avoid obscuring the main point of the investigation. Sometimes, you will be asked to carry out an
investigation in order to dispute or challenge a claim that has been made against a generally accepted
scientific phenomenon. If that is the case, state the currently held theory and the claim that has been made.
Your introduction must state the aim of your investigation. The integration of the aim into the introduction
allows for the smooth transition from general information to the specific goal of the investigation.
Method:
Present a clear, concise and step-by-step description of the apparatus, techniques and procedures used. List
and name the apparatus with brief descriptions of the main parts as well as the functions thereof. A neatly
sketched and labelled diagram of the experimental set-up is essential and can save you paragraphs of tedious
written descriptions. Briefly describe the procedures that you followed in the investigation. Wherever
appropriate, give a reason for each step you took in a procedure. Sometimes more than one section is
required for this material. For example, if two techniques were used, i.e. Technique No.1, Technique No.2,
etc. then a brief explanation of each of these techniques must be given. Do not omit any significant steps. A
description of method helps you to recall the problems that were associated with experimental procedures
such as precision of instruments, strengths and weaknesses of certain techniques, recording of observations
etc. This will help you when you have to summarise your conclusions and recommendations at the end of
your report.
Results and discussion
The main part of this section are your tables of results and graphs. You must have a consistent way of
recording your observations and calculations. Data are normally summarised and displayed in tables and
graphs. Each table and graph is usually referenced by a number and should be numbered in sequence, e.g.
Table 1, Table 2, Figure 1, Figure 2 etc. Each table is accompanied by a title and each graph by a caption which
describes the purpose for which it has been presented. (e.g. “Table 1: Measurements of the width of the
cylinder” and “Graph 3: To determine the viscosity of the sample of oil”).

8
Tables and figures must be referred to in the text, e.g. “The apparatus was arranged as shown in Figure 1”;
“The data gathered were recorded as shown in Table 1 below” or “The data in Table 1 were used to plot the
graphs in Figure 1, 2 and 3.” These brief statements help to link the different parts of the report.
Then examine and extract important aspects of the data and use these to explain various relationships or
determine your measurement results. For example, what is the shape of your graph and what does it suggest
in terms of the relationship between the variables? If it is a straight line, what is the value of the slope?
Remember to quote your results with the appropriate significant figures and the corresponding uncertainties.
How does an expected value compare with your own and what reasons can you give for this? You may give
tentative explanations for your data but be careful not to mix facts with opinions.
Conclusion and recommendations:
This is normally a section in which you say what the investigation has shown and to what extent the problem
or claim stated in the introduction has been resolved. Remember that any conclusions must be supported by
evidence from your data. Always quote any final results, together with their uncertainties in your conclusion.
Avoid making vague statements such as “This was a successful experiment.” You may also need to discuss
sources of uncertainty and any improvements that could be made to the apparatus (and measurements).
Again, avoid meaningless phrases such as “it was caused by human error.”
Writing a report allows you to reflect critically on the whole experiment and check your understanding of the
purpose of the investigation as well as produce an accurate record of it. Note that an physics practical is not a
set of procedures designed to reproduce some “correct” answer. It is a problem that has been posed that
requires an experimental solution which may include making measurements, implementing different
procedures and techniques and then the formulation of a suitable report.

Whether you graduate and leave with a B.Sc. to work in industry or whether you stay on at
university to become a post-graduate student, you will find that report writing will remain as
one of the most important activities in your career.

Scientific style
Very often, in reports of this kind, writers prefer to use the passive construction or impersonal style to report
procedures followed in conducting experiments, by writing, for example:
“Five measurements were taken,” instead of “I / We took five measurements.”
Both styles are acceptable. As you do most of your practicals in groups, you are likely to visualise what you did
as a group and report it as a group activity in which case the personal pronoun “we” is appropriate. However,
once you have chosen a style of writing, then you must use it throughout your report and not switch back and
forth between the two.

Marking the report

When marking your reports, feedback will be given that is relevant to your individual report. All the feedback
you get is intended to help you consider ways of improving your report. Marks for the report will be given for
your data collection and processing (which includes your method, tables, graphs and calculations) as well as
the overall coherence of the report as a piece of writing. The important thing is that you understand the mark
that you receive and speak to someone about it if not. The value of this process cannot be underestimated.
Make sure you follow up with the demonstrator who marked your report in the following practical so you
know how to improve for next time.

The laboratory report checklist

When preparing your report, use the following checklist to see whether or not you have included everything
correctly.

9
 Guide to Reporting and Measurement: January 2018

Introduction and aim:

✓ Have I written my name, my partners’ names and my course on the front of my report book?
✓ Does my report have a title and a date?
✓ Does my aim clearly outline the purpose(s) of my experiment (in one or two sentences)?
✓ Have I clearly outlined the theory involved in my investigation?
Apparatus:
✓ Have I drawn the experimental set-up (with a diagram)?
✓ Is my diagram labelled and does it have a heading?
Method:
✓ Have I described how I went about my experiment (i.e. what I did)?
✓ Have I clearly described the significant steps in the procedure (including the analysis on my data)?
✓ Have I explained why certain aspects of the procedure were undertaken?
Data / Results:
✓ Have I recorded all relevant measurements?
✓ Do all my tables have titles?
✓ Do all the columns in my tables have headings with units?
✓ Are all my data recorded correctly?
Graphs:
✓ Does my graph have a title which should state why I plotted my graph?
✓ Have I used a decent scale so my graph fills the whole page?
✓ Have I labelled the axes, including the correct units?
✓ Have I used a  for the data points and not a  (blob)
✓ Have I used pencil for my graph (and not pen)?
✓ Have I drawn the best straight line through my data points? (if necessary)
✓ Have I shown how I got the slope from my line? (if necessary)
✓ Have I recorded the slope and intercept (with uncertainties) correctly?
Analysis and discussion:
✓ Have I described and explained important relationships revealed by the data (and graphs)?
✓ Have I clearly set out my calculations?
✓ Are my calculations in the correct section of the report?
✓ Have I used SI units throughout?
✓ Have I shown how I calculated the uncertainties in the results?
✓ Is my final result presented correctly with the appropriate significant figures?
✓ Have I described and explained the results of my calculations?
Conclusion and recommendations:
✓ Does my conclusion refer back to my aims for doing the experiment?
✓ Have I fulfilled all my aims for doing the experiment?
✓ Have I quoted all my results with their uncertainties?
✓ Have I recommended how this experiment may be improved?

10
 A3 Drawing tables and graphs
Tables are a way of organising your data in a structured way while you are working in the laboratory as well as
a way of presenting your data in a clear way when you write your report. A graph is an extremely useful way
to both present readings of two variables that vary as a function of each other, as well as to analyse the data
in order to extract further information. This section describes what we call the “general rules” for preparing
tables and graphs.

Deciding how many readings to take

Say that you are recording readings of a quantity in an experiment that varies as a function of some other
quantity. Since it is impossible to take an infinite number of readings, you will need to usually consider the
following:
i. Over what range do you want to record data? The range is the difference between the largest and
smallest reading in the series.
ii. How many readings between the smallest and largest reading do you want to record, and how will
these readings be spaced apart?
Answering these questions before you start to take readings allows to plan your experiment properly and
therefore use your time most effectively.

Guidelines for drawing up tables of data

Here are a few guidelines when drawing up a table.
i. Plan your table, especially how many columns and rows you will need. Try to anticipate what
readings you will take. Remember to include columns for the results of any calculations. For
example, you may sometimes need to calculate the average of a number of repeated readings.
ii. Each table must have a title which should reflect your reason for recording the data. If you use more
than one table in your report, then number them Table 1, Table 2, etc.
iii. Each column must have a heading. Units should be included with the heading and not written
alongside each reading in the table.
iv. Record your data carefully. Remember, for example, that writing 36.0 cm and
36 cm is not the same. In the first case you are recording a reading to the nearest millimetre and in
the second case you only measured to the nearest centimetre.
An example of a table:
Table 1: Data of the extension of a spring due to an applied force Title
Headings with units Applied force, 𝐹 (N) Extension of spring, 𝑥 (m) 𝑥̅ (m)
0.135 0.039 0.038 0.039 0.030
0.270 0.076 0.075 0.077 0.076
0.406 0.114 0.112 0.113 0.113
0.541 0.150 0.149 0.152 0.150
0.676 0.185 0.186 0.186 0.186

Data Calculations

11
 Guide to Reporting and Measurement: January 2018

Guidelines for plotting a graph

i. Use a pencil.
ii. Your graph must have a caption (or title) which should describe why you plotted the graph.
iii. Each axis should be labelled with the name of the variable and the units.
iv. Choose appropriate scales on the axes so that the graph will not be too small on the page, but will
cover a fair portion of the page in each direction.
v. Use a  or  for the data points and not a  (blob).
vi. Decide whether or not each axis should start from zero (it is not always necessary to show the origin).
When do you need to draw the axes from zero?
vii. Axes should be marked in factors of 1, 2, 5, or these times a power of ten. Other factors such as 3 or 4
usually make your scales difficult to read between divisions.
viii. The line or curve that you draw through your data should be a reasonable “best fit” so as to model
the trend of the experimental points. Your graph should not join the points.
See below for an example of a straight line graph:

Figure 1: Graph of extension of the spring versus applied force in order to

determine the spring constant of a spring.
0.300
Extension
(m)
0.200 


0.100


0.000
0.000 0.200 0.400 0.600 0.800
Force (N)

Interpretation of graphs
One of the aims in investigating physical phenomena is to establish the relationship between the variables
that are being measured. For example, if we are investigating an object that is experiencing uniform
acceleration, 𝑎, then we will find the function describing the relationship between velocity, 𝑣, and time, 𝑡, to
be a straight line of the form,
𝑦 = 𝑚𝑥 + 𝑐;
an example of which is shown in graph (a) of Figure 2. In this case,
𝑣 = 𝑢 + 𝑎𝑡 ,
where 𝑢 is the initial velocity.
On the other hand, the relationship between the position, r, and the time, t, will be of the form,
𝑦 = 𝑎𝑥 2 + 𝑏𝑥 + 𝑐,
i.e. a parabola as shown in graph (b) of Figure 2. In this example we will find,
1
𝑟(𝑡) = 𝑟0 + 𝑢𝑡 + ,
2𝑎𝑡 2

12
where 𝑟0 is the initial position. Some other functions that you will encounter are the hyperbola, which is
shown in graph (c) of Figure 2 and of the form,
𝑥𝑦 = 𝑐,
and the exponential function which is shown in graph (d) of Figure 2 and of the form,
𝑦 = 𝑎𝑒 𝑏𝑥 .
Note that we say that “𝑦 is proportional to 𝑥” if a graph of 𝑦 versus 𝑥 yields a straight line through the origin.
i.e. 𝑦 = 𝑚𝑥 + 𝑐 with 𝑐 = 0. Sometimes “directly proportional to” is used which means the same thing.
However, if 𝑐  0, then we cannot say that “𝑦 is proportional to 𝑥”, but can only say that “𝑦 is linearly related
to 𝑥”.



y (a) y (b)
 







 Straight line Parabola



 𝑦 = 𝑚𝑥 + 𝑐  𝑦 = 𝑎𝑥 2 + 𝑏𝑥 + 𝑐


x x
y y
 (c) (d)



Hyperbola Exponential
𝑥𝑦 = 𝑐  𝑦 = 𝑎𝑒 𝑏𝑥








 


x x

Figure 2: Different types of graphs that you may encounter.

13
 Guide to Reporting and Measurement: January 2018

A4 Significant figures
What are significant figures?
The number of figures that convey meaningful information in recorded data are known as the “number of
significant figures”.
Significant figures in a number are:
 All non-zero digits (e.g., 54 has two significant digits while 456.78 has five significant digits).
 All zeros appearing anywhere between two non-zero digits (e.g., 302.05 has five significant digits).
 All trailing zeros, where the number has a decimal point (e.g., 36.65000 has seven significant digits).
 Leading zeros are not significant (e.g., 0.0003 has only one significant figure).
Do not confuse significant figures with decimal places. The number of decimal places refers to the number of
figures after the decimal point, e.g., 46.320 mm has five significant figures, but three decimal places; while
0.0040 mm has two significant figures, but four decimal places.
Recording significant figures in readings and calculations
When taking readings, the number of significant figures recorded is determined by the precision of the
instrument. For example, if the scale on a particular instrument makes it possible to read a mass to 1/10th of a
gram, then it is incorrect to record the reading of some mass to 1/100th of a gram, i.e., in such a case the
recording of a mass of say 12.4 g would be correct, but recording that same mass as 12.40 g using the same
instrument would be incorrect.
Similarly, when using a typical wooden laboratory metre rule to measure distance, you can only take readings
to the nearest millimetre. For example, you may record the reading of a distance as say 127 mm, but to
record a reading of 127.0 mm will be unreasonable because that degree of precision (to 0.1 mm) cannot be
achieved when using a typical laboratory meter rule. (Consider the problem of parallax.)

14
SECTION B – MEASUREMENT UNCERTAINTY

B1 Overview
The purpose of undertaking a measurement in science is to provide knowledge about some physical quantity.
The physical quantity that we may wish to investigate is called the measurand, e.g., the measurement may be
the air pressure in a soccer ball, the temperature of the air in a room, or the voltage across a resistor.
It is important to realise that the value of a measurand can never be determined with absolute certainty. The
best that can be achieved through a process of measurement is to improve our knowledge about some
specified measurand. We cannot think of a measurand as having some “true” or “exact” value that can
actually be found; we have to think of the knowledge that we have of any measurand always as being limited
and incomplete.
The following framework for measurement uncertainty is based on a probabilistic approach and the relevant
standards are expressed in ISO/IEC 17025. These standards are clarified in the associated Guide to the
Expression of Uncertainty in Measurement (GUM) as well as in the ISO International Vocabulary of Basic and
General Terms in Metrology (VIM). They are also documented in the NIST Technical Note 1297 of 1994.
Note that the term “error” is often used quite loosely in conversation in laboratory work but it is important to
realise that the word “error” is not synonymous with “uncertainty”. The term “uncertainty” is clearly defined,
while use of “error” may be misleading, so the use of the word “error” in the context of laboratory work is
discouraged.

Decide on the best approximation of the measurand

Having taken the readings, 𝑥1 or (𝑥1 , 𝑥2 , 𝑥3 , …. 𝑥𝑛 for n readings) as the case may be, the next step is to
decide on the best approximation of the value of the measurand.
In the case where only one reading has been taken, then that reading is the best approximation of the value
of the measurand, e.g., 𝑥. However, where more than one reading has been taken, then the best
approximation of the value of the measurand is the mean (average) of those readings, e.g., 𝑥̅ .

List all the possible sources of uncertainty

There may be many sources of uncertainty to be taken into account in any one measurement. These sources
include, for example: a) the manufacturer’s rating of the instrument used, b) the reading of the scale of the
instrument, c) corrections that might need to be included as a result of ambient conditions, and so on…
Which of these sources of uncertainty are to be taken into account in any particular measurement is based
mainly on experience and skill in measurement, but the important first step is to identify and list them.

Evaluate the uncertainties that have been identified.

Having identified the relevant sources of uncertainty in the measurement of 𝑥, we can evaluate the value of
each uncertainty. The key value of the uncertainty in the measurement is known as the standard uncertainty
and the symbol is 𝑢(𝑥). Subscripts are used to indicate difference sources; for example, 𝑢(𝑥rating ) and
𝑢(𝑥reading ).
The way in which we evaluate the uncertainty in our knowledge of a measurand has its roots in a method of
statistical inference based on the Bayesian theorem; and the functions used to describe how the uncertainty
in our knowledge of a particular measurand is formulated are known as probability density functions (pdfs).
However, in this course we will not delve into the statistical underpinnings of uncertainty evaluation, nor will
we be using pdfs in a rigorous way, but we will refer to the relevant pdf as appropriate and will simply use the
method as required.

15
 Guide to Reporting and Measurement: January 2018

There are two ways to evaluate the standard uncertainty, a TYPE A evaluation and a TYPE B evaluation; and
the method you will use depends on the nature of the measurement.
 If a number of readings are available, in which there is scatter, then a statistical, TYPE A, evaluation of
the uncertainty should be used.
 All other evaluations of uncertainty are TYPE B evaluations. TYPE B evaluations may be applied to
many sources of uncertainty, e.g., the rating of the instrument, the reading of a scale, the correction
factors that may need to be applied to the measurement, etc.
The TYPE A and TYPE B evaluation methods are discussed in detail in the sections that follow.

Combine the standard uncertainties to find the combined standard uncertainty.

Depending on the nature of the measurement it may be necessary to combine the standard uncertainties,
(𝑥rating ), 𝑢(𝑥reading ), etc., to give a combined standard uncertainty.

Quote the result of the measurement

The final step is to quote the result of the measurement: e.g., 𝑥 ± 𝑘. 𝑢𝑐 (𝑥), stating the relevant coverage
probability (also referred to as the ‘level of confidence’).
See D1 Quoting a result for more information on reporting the result of the measurement.

16
 B2 TYPE A evaluation of uncertainty
In a TYPE A evaluation of the uncertainty, a statistical method is used to infer the value of the uncertainty
associated with a measurement by quantifying the “scatter”, or the “spread” in the values of a set of data.
There are two commonly used cases:

1) The data are a set of readings of the form: 𝑥1 , 𝑥2 … . 𝑥𝑛

A histogram of the readings shows a distribution around some mean value 𝑥̅ that can be represented by a
symmetrical bell-shaped curve (referred to as a “Gaussian”) where: 1 −[
(𝑥−𝑥̅ )2
]
𝑦= 𝑒 2𝜎2
𝜎√2𝜋

Figure 3: Normal or ‘Gaussian’ distribution function.

When the relevant set of readings, 𝑥1 , 𝑥2 … . 𝑥𝑛 , for n readings are plotted as a histogram, and the histogram
allows one to fit a curve similar in shape to a normal distribution, then the best approximation of 𝑥 is x , and
𝜎
the standard uncertainty 𝑢(𝑥) is given by the experimental standard deviation of the mean, .
√𝑛

2) The data are a set of readings of the form: (𝑥1 , 𝑦1 ), (𝑥2 , 𝑦2 ) … . (𝑥𝑛 , 𝑦𝑛 )
In this case the function that is used to analyse the data is
‘linearised’ (details to be covered in E2 Linearising
equations) and the data pairs, (𝑥𝑖 , 𝑦𝑖 ) i = 1, 2, …n are
plotted as a straight line in the form 𝑦 = 𝑚𝑥 + 𝑐.
A technique known as a ‘least squares fit’ is used to fit a
line to the data and then the best approximation and the
standard uncertainty may be derived from the gradient
𝑚 ± 𝑢(𝑚) and the intercept 𝑐 ± 𝑢(𝑐)of the fitted line.

Formulae for a TYPE A evaluation of uncertainty

Consider an experiment in which a set of readings has been taken, where 𝑥1 , 𝑥2 … . 𝑥𝑛 , and n is the number of
readings and 𝑥𝑖 is the i th reading. If there is a scatter in the values of these readings, then it can be reasonably
assumed that the distribution of the readings ‘fits’ a Gaussian probability density function.
Using this assumption*, there are three values required to calculate the uncertainty in the measurement:
 the mean 𝒙̅ (sometimes the symbol µ is also used for the mean)
 the standard deviation σ, and
 the experimental standard deviation of the mean which, when the distribution of the values of the
mean of the readings is normal, is the standard uncertainty 𝒖(𝒙).

17
 Guide to Reporting and Measurement: January 2018

The following equations are used to calculate the parameters used to determine the uncertainty associated
with the measurement:

𝑛
1
Mean of the data: 𝑥̅ = ∑ 𝑥𝑖 … (B2-1)
𝑛
𝑖=1

𝑛
1
Standard deviation: 𝜎=√ ∑(𝑥𝑖 − 𝑥̅ )2 … (B2-2)
(𝑛 − 1)
𝑖=1

𝜎
Standard uncertainty: 𝑢(𝑥) = … (B2-3)
√𝑛

Note:

1) *The assumption that the distribution of the values of the readings is normal validates the statement
that “the experimental standard deviation of the mean is the standard uncertainty 𝑢(𝑥)” for the
majority of cases. However, it is important to realise that a different approach may be required where
the distribution of values is not ‘normal’.
2) Because a Gaussian distribution was assumed, the probability that the measurand lies within one
standard uncertainty of the best approximation, is 68%.

Example: TYPE A evaluation of uncertainty associated with a set of data

Consider an example in which the measurand is the period T of an oscillating pendulum. A stopwatch has
been used to take ten readings (n = 10) of the time taken for twenty oscillations of the pendulum. It is
assumed that the distribution of the values in this set of data can be modelled by a Gaussian distribution.
Table B2.1: Data recorded using digital stopwatch.

Time for 20 oscillations, T20 (s) Period, T = T20/20 (s)

19.43 21.65 0.972 1.083
20.49 20.82 1.025 1.041
20.76 19.77 1.038 0.989
20.63 20.39 1.032 1.020
20.56 19.02 1.028 0.951
Using Eq. (1), the mean, 𝑇̅ = 1.0176 s. This is the best approximation of the measurand.
Using Eq. (2), the standard deviation σ is 0.0379 s. However, this value is not the standard uncertainty. For
that we still need to calculate the experimental standard deviation of the mean, which is the standard
uncertainty in this case.
Using Eq. (3), the standard uncertainty 𝑢(𝑇) = 0.0119 s.
The result of this measurement, the period T of the pendulum, may be quoted as:
T = (1.018 ± 0.012) s, and since this is to one standard uncertainty - and a Gaussian pdf was used - the
coverage probability is 68%.
Note: The result may be quoted to a higher coverage probability (a greater level of confidence) by using an
expanded uncertainty, see D1 Quoting a result.

18
 B3 TYPE B evaluation of uncertainty
A TYPE B evaluation of uncertainty is any evaluation of an uncertainty that is not a TYPE A evaluation. So, for
example, if the uncertainty as a result of the rating of the instrument has to be evaluated, or if only one
reading is available, then a non-statistical TYPE B evaluation of uncertainty is performed.
The methods of TYPE B evaluation of uncertainty will be presented by way of examples.

Example 1: Uncertainty associated with the rating of an instrument

Manufacturers of measuring instruments will state the information required to evaluate the uncertainty
associated with the rating of the instrument, e.g. 𝑢(𝑥rating ), by specifying the rating as:
a) a percentage of the maximum value that can be displayed, or
b) a percentage of the full scale deflection (FSD), or
c) simply by stating that the rating is a percentage of whatever is being displayed on the instrument.
In the first example we consider that the measurand is a pressure p and the instrument being used to take the
readings of the pressure in this case is a typical bourdon tube gauge.

Figure 4: Reading on a bourdon tube pressure gauge.

Just by looking at the position of the needle on the gauge we can see that the best approximation of the
measurand is: p = 34 kPa. (There is only one reading.)
Now we want to evaluate the uncertainty associated with the manufacturer’s rating of the instrument,
𝑢(𝑝𝑟𝑎𝑡𝑖𝑛𝑔 ). The reason for selecting this source is because the bourdon gauge is a simple instrument and so
the uncertainty due to the rating of the instrument far outweighs the contribution of all the other applicable
sources of uncertainty. Typically, for a bourdon tube gauge, the uncertainty in the rating of an instrument of
the type shown here is ±2.5% of the full scale deflection (FSD) of the instrument; and in this case, FSD is 160
kPa. The standard uncertainty (due to the rating) associated with this instrument will be:

2.5
𝑢(𝑝𝑟𝑎𝑡𝑖𝑛𝑔 ) = 160 kPa × = 4.0 kPa … (B3-1)
100

And so the result of the measurement of the pressure shown on this instrument will be quoted as:
𝑝 ± 𝑢(𝑝) = (34.0 ± 4.0) kPa.
In every measurement, the instrument manufacturer’s specification needs to be checked to determine the
instrument rating. Caution: check whether the manufacturer’s rating has been given to one or more standard
deviations; as this has to be taken into account when you combine the uncertainties, see D1 Quoting a result.

19
 Guide to Reporting and Measurement: January 2018

Example 2: Uncertainty associated with reading a digital display

In this example, the measurand is a temperature T and the instrument chosen to take readings of the
temperature is an electronic thermometer that has a 3-digit digital display. The reading you see on the
thermometer is shown in Figure 5.

Figure 5: Reading on a digital thermometer.

By inspection we can see that the best approximation of the measurand is: 𝑇1 = 25.3 °C. (There is only one
reading.)
Now we want to evaluate the uncertainty associated with the reading of the display, 𝑢(𝑇read ). This
uncertainty arises because even though the display shows 25.3 °C, any actual temperature between 25.25 °C
and 25.35 °C would result in the same display on the instrument when rounded to 3 digits.
The evaluation of 𝑢(𝑇read ) proceeds as follows:
1) The only information we have about the temperature is that it is between a lower bound, 𝑇𝑙𝑜𝑤𝑒𝑟 =
25.25 °C and an upper bound, 𝑇upper = 25.35 °C.
2) We make the assumption that it is equally probable for the temperature T to be any value between
𝑇lower and 𝑇upper .
3) Having assumed that it is equally probable that the value of T lies anywhere in a symmetric interval
around 25.30 °C, this uncertainty evaluation can be modelled on a uniform or rectangular probability
density function, see Figure 6, and so the standard uncertainty 𝑢(𝑇read ) can be evaluated as:

(𝑇upper − 𝑇lower ) (25.35 − 25.25) °C

𝑢(𝑇𝑟𝑒𝑎𝑑 ) = = = 0.029 °C … (B3-2)
2√3 2√3

The reason for dividing by √2 in equation B3-2 is because this is the factor applicable to the assumptions
regarding a rectangular pdf.

Figure 6: Rectangular probability density function.

Note that the total area under the pdf is equal to 1, meaning that there is a 100% probability of the measurand
being in the ± a interval. Moreover, the grey shaded area under the rectangular pdf has an area of 0.58, so
there is a 58% probability that the measurand is in the ± u interval. Therefore, because a rectangular distribution
was assumed, the coverage probability that the measurand lies within one standard uncertainty, i.e., within ±
𝑢(𝑇) of the best approximation T, is 58%.

20
 Example 3: Uncertainty associated with reading an analogue display
In this example, the measurand is a temperature T but now the instrument chosen to take readings of the
temperature has an analogue display. The reading you see on the thermometer is shown in Figure 7.
The best approximation of the measurand is 𝑇1 = 75 °C (there is only one reading).
In this example the only source of uncertainty that we will take into account is the uncertainty associated with
the reading of the dial of the instrument, 𝑢(𝑇read ). This uncertainty arises because the instrument really tells
us is that the temperature is between 70 °C and 80 °C, and is most probably 75 °C.

Figure 7: Reading on an analogue thermometer.

The evaluation of 𝑢(𝑇read ) proceeds as follows:

1) As was the case in Example 2, we judge that the temperature is between a lower bound 𝑇lower = 70 °C
and an upper bound 𝑇upper = 80 °C.
2) However, in this case it is not equally probable that temperature T will be any value between 𝑇lower
and 𝑇upper . The temperature T is most likely to be a value ‘close to’ 75 °C, which we can say as the
needle is approximately half way between 70 and 80 °C. The choice of the upper and lower limits is
arbitrary, and tends to be a pair positioned symmetrically around the most likely value, and aligned
with tick marks of the instrument.
3) Having assumed that it is most probable that the value of T lies at the centre of a symmetric interval
around 75 °C, this uncertainty evaluation can be modelled on a triangular probability density function,
see Figure 8, and so the standard uncertainty 𝑢(𝑇read ) can be evaluated as:

(𝑇upper − 𝑇lower ) (80 − 70) °C

𝑢(𝑇read ) = = = 2.1 °C … (B3-3)
2√6 2√6

The reason for dividing by √6 in equation B3-3 is because this is the factor applicable to the assumptions
made regarding a triangular pdf.
Note that the total area under the pdf is equal to 1, meaning that there is a 100% probability of the
measurand being in the ± a interval. Moreover, the grey shaded area under the triangular pdf has an area of
0.65, so there is a 65% probability that the measurand is in the ± u interval. Therefore, because a triangular
distribution was assumed, the coverage probability that the measurand lies within one standard uncertainty,
i.e., within ± 𝑢(𝑇) of the best approximation T, is 65%.

Figure 8: Triangular probability density function.

21
 Guide to Reporting and Measurement: January 2018

B4 Combination of uncertainty in one measurement

In the previous three examples the uncertainty associated with only one possible source of uncertainty was
evaluated in each example. However, there are many possible sources of uncertainty that may have to be
taken into account in any one measurement and once these have been identified and evaluated the
combined standard uncertainty in the measurement has to be calculated.

General formula for combining uncertainties

Assume that the measurand is some quantity k and that some number of sources of uncertainty have been
identified and designated as: 𝑢(𝑘1 ), 𝑢(𝑘2 ), 𝑢(𝑘3 ), 𝑢(𝑘4 ) …
Once the uncertainties associated with each of these sources has been evaluated, the values of the standard
uncertainties may be combined, in quadrature, to find the combined standard uncertainty 𝑢(𝑘), as follows:

𝑢(𝑘) = √[𝑢(𝑘1 )]2 + [𝑢(𝑘2 )]2 + [𝑢(𝑘3 )]2 + [𝑢(𝑘4 )]2 + ⋯ … (B4-1)

It does not matter whether the uncertainty associated with the individual sources has been determined by
means of a TYPE A or a TYPE B evaluation, they can all be combined using equation B4-1. Before you combine
uncertainties, make sure that they have all been adjusted to relate to the one (or the same) standard
uncertainty.

Example 1: uncertainty associated with using an analogue voltmeter

In this example, the measurand is the voltage V across a torch battery. An analogue voltmeter is chosen to
determine the voltage and when the leads of the voltmeter are connected to the battery, the reading shown
in Figure 9 appears on the display of the voltmeter:

Figure 9: Reading on an analogue voltmeter display.

The displayed reading is the best approximation of the measurand, in this case 1.55 V.
Consider that in this measurement there are two sources of uncertainty to be evaluated:
1) the uncertainty associated with the rating of the instrument, 𝑢(𝑉rating ) and
2) the uncertainty associated with reading the display, 𝑢(𝑉read ).
The typical rating for a moving coil instrument of this kind is 2% of the full scale deflection (FSD).
2
So in this case the standard uncertainty, 𝑢(𝑉rating ) = 3.0 V × = 0.060 V.
100

To evaluate the uncertainty in the reading, consider that the lower and upper bounds of the value of the
measurand are 1.5 V and 1.6 V and that the evaluation of this uncertainty may be modelled on a triangular
pdf, so

(𝑉upper − 𝑉lower ) (1.6 − 1.5) V

𝑢(𝑉read ) = = = 0.021 V.
2√6 2√6

22
The standard uncertainties are combined:

2
𝑢(𝑉) = √[𝑢(𝑉rating )] + [𝑢(𝑉read )]2 = √(0.060 V)2 + (0.021 V)2 = 0.064 V

Finally, the battery voltage can be quoted as V = (1.550 ± 0.064) V, to one standard uncertainty. Given that the
standard uncertainty was obtained by using a triangular probability density function, the coverage probability
(level of confidence) in this result is 65%.
The oscilloscopes in PHYLAB1 have the following ratings, to one standard uncertainty:
 ±5% on the x axis (for time-base between 0.1 μs and 50 ms per division),
 ±3% on the y axis (for voltage between 5 mV and 5 V per division).

Example 2: measurement of VOLTAGE using an oscilloscope

Consider an example in which the measurand is the peak-to-peak voltage V of a sine wave produced by a
signal generator. You connect the signal generator to an oscilloscope and you see the wave shown in Figure
10:

Figure 10: Reading a voltage on an oscilloscope.

After careful consideration, you decide that the peak-to-peak voltage spans 5.8 divisions and so the best
estimate of V is 5.8 V (5.8 Divisions × 1.0 Volt/Division).
Now there are three sources of uncertainty to be evaluated:
1) the uncertainty associated with the instrument rating,
2) the uncertainty associated with the reading at the top peak on the display, and
3) the uncertainty associated with the reading at the bottom peak on the display.
The oscilloscopes in PHYLAB1 have a rating of ±3% on the y axis to one standard uncertainty (for voltage
between 5 mV and 5 V per division). Therefore the standard uncertainty associated with the rating is,

3
𝑢(𝑉rating ) = 5.8 V × = 0.17 V.
100

To evaluate the uncertainty associated with the reading of the display, consider that the lower and upper
bounds of every reading will be 0.1 of a division on either side of the best approximation anywhere on the
display and that the evaluation of this uncertainty may be modelled on a triangular pdf. We need to multiply
the uncertainty in divisions by the sensitivity scale of the oscilloscope settings to achieve an uncertainty in
volts:

23
 Guide to Reporting and Measurement: January 2018

(𝑉upper − 𝑉lower ) (0.1 − (−0.1)) DIV V

𝑢(𝑉read ) = = × 1.0 = 0.041 V.
2√6 2√6 DIV

Finally, these three standard uncertainties are combined to give the combined standard uncertainty
associated with the measurement of V. Note the uncertainty in reading the top and bottom peaks on the
display are equivalent:

2 2 2
𝑢(𝑉) = √[𝑢(𝑉rating )] + [𝑢(𝑉read,top )] + [𝑢(𝑉read,bottom )] = √(0.17 V)2 + 2(0.041 V)2 = 0.18 V.

The result of this measurement can be quoted as:

Voltage, V = (5.80 ± 0.18) V, to one standard uncertainty.
Given that the standard uncertainty was obtained by using a triangular probability density function, the
coverage probability (level of confidence) in this result is 65%.

Example 3: measurement of PERIOD using an oscilloscope

Next consider an example in which the measurand is the period T of a sine wave produced by a signal
generator. You connect the signal generator to an oscilloscope and you see on the screen the wave shown in
Figure 11:

Figure 11: Reading a period on an oscilloscope.

After careful consideration, you decide that the period spans 7.6 divisions, and so the best approximation of T
is 3.8 ms (i.e. 7.6 Divisions × 0.5 ms/Division).
Again there are three sources of uncertainty to be evaluated:
1) the uncertainty associated with the instrument rating,
2) the uncertainty associated with the reading on the left of the display, and
3) the uncertainty associated with the reading at the right of the display.
The oscilloscopes in PHYLAB1 have a rating of ±5% on the x axis to one standard uncertainty (for time-base
between 0.1 μs and 50 ms per division). Therefore, the standard uncertainty associated with the rating is,

5
𝑢(𝑇rating ) = 3.8 ms × = 0.19 ms.
100

To evaluate the uncertainty associated with the reading of the display, consider that the lower and upper
bounds of every reading will be 0.1 of a division on either side of the best approximation anywhere on the
display and that the evaluation of this uncertainty may be modelled on a triangular pdf. We need to multiply
the uncertainty in divisions by the sensitivity scale of the oscilloscope settings to achieve an uncertainty in
volts:

24
 (𝑇upper − 𝑇lower ) (0.1 − (−0.1)) DIV ms
𝑢(𝑇𝑟𝑒𝑎𝑑 ) = = × 0.5 = 0.021 ms.
2√6 2√6 DIV

As with the measurement of the voltage, these standard uncertainties are combined to give the combined
standard uncertainty associated with the measurement of T:

2 2 2
𝑢(𝑇) = √[𝑢(𝑇rating )] + [𝑢(𝑇read,left )] + [𝑢(𝑇read,right )] = √(0.19 ms)2 + 2(0.021 ms)2 = 0.19 ms.

Note that as 𝑢(𝑇rating ) is considerably larger than 𝑢(𝑇read ) by an order of magnitude larger, it dominates the
combined uncertainty, 𝑢(𝑇).
The result of this measurement can be quoted as:
Period T = (3.80 ± 0.19) ms, to one standard uncertainty.
And given that the standard uncertainty was obtained by using a triangular probability density function, the
coverage probability (level of confidence) in this result is 65%.

25
 Guide to Reporting and Measurement: January 2018

SECTION C – CALCULATIONS WITH UNCERTAINTIES

C1 Calculation of uncertainty across multiple measurements

Often, having completed two or more measurements where the results of those measurements are say 𝑥 ±
𝑢(𝑥) and 𝑦 ± 𝑢(𝑦), you need to use these results to calculate some other value, say 𝑧 and its associated
uncertainty 𝑢(𝑧). Caution: you cannot perform the same operation on the uncertainties as you do on the
variables (e.g. in the case of 𝑧 = 𝑥 + 𝑦, 𝑢(𝑧) ≠ 𝑢(𝑥) + 𝑢(𝑦)).
The formulae required to calculate or propagate uncertainties are given in Table C1-1. We will not derive
these equations, as this requires partial derivatives which you may not have met yet in your mathematics
course.
Table C1-1: Formulae used in the calculation of uncertainties.
Form of equation from which Formula for calculating
result R is calculated the standard uncertainty 𝑢(𝑅)

Sum of variables [𝑢(𝑅)]2 = [𝑎 𝑢(𝐴)]2 + [𝑏 𝑢(𝐵)]2

𝑅 =𝑎𝐴 ±𝑏𝐵±𝑐 or
Coefficients a, b & c are constants (numbers
with zero uncertainty) 𝑢(𝑅) = √[𝑎 𝑢(𝐴)]2 + [𝑏 𝑢(𝐵)]2

2 2 2
𝑢(𝑅) 𝑢(𝐴) 𝑢(𝐵)
[ ] = [𝑎 ] + [𝑏 ]
Product of variables 𝑅 𝐴 𝐵
𝑅 = 𝑐 𝐴𝑎 𝐵𝑏 or
Coefficients a, b & c are constants (numbers 2 2
with zero uncertainty) 𝑢(𝐴) 𝑢(𝐵)
𝑢(𝑅) = 𝑅√[𝑎 ] + [𝑏 ]
𝐴 𝐵

Correlated variables
(consider co-variance if the instrument used more than
once in the same experiment)
𝑢(𝑅) = √𝑎 𝑢(𝐴)
𝑅=𝑎𝐴
Coefficient a is a constant (numbers
with zero uncertainty).

Note: these equations are three results from a general function for the propagation of uncertainties. To find out more, consult the full
version of this guide at the UCT Physics website, “Measurement Manual”.

26
 C2 Examples of uncertainty calculations

Example 1: Uncertainty associated with reading a metre stick

Consider a pendulum (see Figure 12), where the effective length is the difference between the position, 𝑥ring ,
of the centre of the ring relative to the metre stick at one end, and the position, 𝑥ball , of the centre of the
mass of the ball relative to the metre stick at the other. So, there are two measurements of position to be
made.

Figure 12: Reading of pendulum length on a metre stick.

In identifying sources of uncertainty, consider that the problem of parallax is a significant contributor to the
uncertainty associated with both measurements, while there is also an uncertainty as to where the centre of
mass of the ball might actually be.
The best approximation for the position of 𝑥ring is 0.0 cm, because the metre stick was lined up that way (we
did not have to start at zero but in this case we did and we read in centimetres as the unit of the instrument).
Using a triangular pdf, the standard uncertainty, 𝑢(𝑥ring ) can be evaluated as:

(𝑥ring,upper − 𝑥ring,lower ) (0.1 − (−0.1)) cm

𝑢(𝑥ring ) = = = 0.041 cm.
2√6 2√6

The position of the centre of the ring relative to the metre stick may be quoted as 𝑥ring = (0.000 ± 0.041) cm,
to one standard uncertainty.
The best approximation of the position of the ball, 𝑥ball is 66.4 cm. Using a triangular pdf, the standard
uncertainty, 𝑢(𝑥ball ) can be evaluated as:

(𝑥ball,upper − 𝑥ball,lower ) (66.6 − 66.2) cm

𝑢(𝑥ring ) = = = 0.082 cm.
2√6 2√6

The position of the centre of the ball relative to the metre stick may be quoted as 𝑥ball = (66.400 ± 0.082) cm,
to one standard uncertainty.
To find the effective length L of the pendulum we need to subtract the position of one from the position of
the other. The best approximation of L is the difference between the two best approximations of the
measurements so:
𝐿 = (𝑥ball − 𝑥ring ) = (66.4 cm − 0.00 cm) = 66.4 cm.
Therefore the best approximation of L is 66.4 cm.

27
 Guide to Reporting and Measurement: January 2018

We combine the two standard uncertainties, 𝑢(𝑥ball ) and 𝑢(𝑥ring ), using the first equation in Table C1-1. The
appropriate formula for the calculation of the combined standard uncertainty is:

2
𝑢(𝐿) = √[𝑢(𝑥ball )]2 + [𝑢(𝑥ring )] = √(0.082 cm)2 + (0.041 cm)2 = 0.092 cm

Finally, the effective length of the pendulum is L = (66.400 ± 0.092) cm, to one standard uncertainty. Given
that the standard uncertainty was obtained by using a triangular probability density function, the coverage
probability (level of confidence) in this result is 65%.

Example 2: Calculating the uncertainty where measurements are correlated.

Consider the situation where a starter pistol is fired and the sound wave reflects off a nearby wall to be heard
by an observer. The total distance, 𝑥, travelled by the sound to return to can be seen as either:
𝑥 = 𝑑1 + 𝑑2 , where 𝑑1 is the distance from the source to the wall, and 𝑑2 is the distance from the
wall to the listener - and you treat the two distances as if these were two
independent measurements -
or
𝑥 = 2𝑑, where the same measurement, d, is used twice. In this case the measurement ‘there’
correlates with the measurement ‘back’, and so a covariance factor of √2 has to be
included in the calculation of 𝑢(𝑥); which means you get the same result as with the
other equation.
Using the first method, where 𝑥 = 𝑑1 + 𝑑2 . By means of the tape measure you determine 𝑑1 = (56.00 ± 0.25)
m, and 𝑑2 = (56.00 ± 0.25) m. Which means that the ‘best approximation’ of the total distance 𝑥 = 𝑑1 + 𝑑2 =
56.00 + 56.00 = 112.00 m.
Now, with reference to Table C1-1, (measurements are being added) you calculate 𝑢(𝑥):

[𝑢(𝑥)]2 = [1 𝑢(𝑑1 )]2 + [1 𝑢(𝑑2 )]2 = [0.25 m]2 + [0.25 m]2 , so 𝑢(𝑥) = 0.3535 m.

Using the second method, where 𝑥 = 2𝑑. By means of the tape measure you determine 𝑑 =
(56.00 ± 0.25 m). Which means that the ‘best approximation’ of the total distance 𝑥 = 2𝑑 = 2(56.00) =
112.00 m.
Now, with reference to Table C1-1, (correlated variables) you calculate 𝑢(𝑥):

𝑢(𝑥) = √2 𝑢(𝑑) = √2 (0.25 m), so 𝑢(𝑥) = 0.3535 m.

The result of either calculation will be quoted as: 𝑥 ± 𝑢(𝑥) = 112.00 ± 0.35 m.

Example 3: Calculating the uncertainty where measurements are multiplied together.

Calculate the speed of sound in air having determined the time taken and the distance covered by a sound.
The results are:
distance, 𝑥 ± 𝑢(𝑥) = (112.00 ± 0.35) m, and
time, 𝑡 ± 𝑢(𝑡) = (0.326 ± 0.018) s.
The calculated ‘best approximation’ of the speed of sound is

𝑥 112.0 m
𝑣= = = 343.3 m s −1 .
𝑡 0.326 s

28
Now, with reference to Table C1-1, (measurements are being multiplied) you calculate 𝑢(𝑥):
2 2 2
𝑢(𝑣) 𝑢(𝑥) 𝑢(𝑡)
[ ] = [1 ] + [−1 ]
𝑣 𝑥 𝑡

2
𝑢(𝑣) 0.36 m 2 0.018 s 2
[ ] = [1 ] + [−1 ]
343.3 m s −1 112.0 m 0.326 s

so 𝑢(𝑣) = 18.99 m s-1.

The result of the calculation will be quoted as: 𝑣 ± 𝑢(𝑣)= (343 ± 19) m s-1.

Example 4: Calculating frequency of an oscillation having measured the period.

Assume the measured period of an oscillation to be 𝑇 ± 𝑢(𝑇) = (2.600 ± 0.055) ms.
Then the calculated ‘best approximation’ of the frequency is
1
𝑓= = 𝑇 −1 = (2.600 × 10−3 )−1 = 384.615 Hz.
𝑇
The uncertainty in this calculated result is
2 2 2
𝑢(𝑓) 𝑢(𝑇) 𝑢(𝑇) 𝑢(𝑇)
[ ] = [− ] = [−1]2 [ ] → 𝑢(𝑓) = 𝑓 From Table C1.1
𝑓 𝑇 𝑇 𝑇
0.055×10−3 s
So 𝑢(𝑓) = (384.615 Hz) = 8.136 Hz.
2.600 ×10−3 s

The result of the calculation will be quoted as: 𝑓 ± 𝑢(𝑓) = (384.6 ± 8.2) Hz.
Note in all these examples, the uncertainty is quoted to two (2) significant figures, and the best approximation
has the same number of decimal places as the uncertainty. See the next section, D1, for more information on
quoting.

The “reductionist approach” to simplify the propagation of uncertainties

Consider Example 3 again in terms of the fractional uncertainty, where the uncertainty is expressed as a
fraction of the best approximation of the measurand (see Section D1):
2 2 2
𝑢(𝑣) 𝑢(𝑥) 𝑢(𝑡)
[ ] = [1 ] + [−1 ]
𝑣 𝑥 𝑡
𝑢(𝑣)
= √[0.0021]2 + [0.055]2 = 0.055
𝑣
Note how the fractional uncertainty in 𝑡 is an order of magnitude larger than the fractional uncertainty in 𝑥.
As we are combining the fractional uncertainties in quadrature (sum of squares), the fractional uncertainty in
𝑣 is dominated by the larger term; the fractional uncertainty in 𝑡. Effectively you only need to consider the
larger component so can approximate the fractional uncertainty in 𝑣,
2 2
𝑢(𝑣) 𝑢(𝑡)
[ ] ≈ [−1 ]
𝑣 𝑡
So you can find, 𝑢(𝑣) = 0.036 𝑣 = 0.055(343.3 ms −1 ) = 18.99 m/s, as before.
Often you can substitute or combine equations to use the reductionist approach to great advantage. You will
have an opportunity to try this out in Practical 5 in Part I, Vibrating String.
Note the reductionist approach is only an appropriate approximation when you have at least one order of
magnitude in difference between different fractional uncertainty components.

29
 Guide to Reporting and Measurement: January 2018

SECTION D – REPORTING RESULTS

D1 Quoting a result
The outcome of a measurement and/or an experiment is the best approximation of the measurand as well as
the standard uncertainty, for example, the result of a measurement of gravitational acceleration 𝑔 may be:
𝑔 = (9.790 ± 0.052) ms −2 .

(best approximation) ± (standard uncertainty) (68% coverage probability)

Note that:
 The result of a measurement is not a single value (not a point), it is an interval of values.
 The greater the uncertainty, the ‘wider’ the interval that quantifies the result of the uncertainty in the
measurement.
 Because there is always some uncertainty in measurement, the width of the interval is never zero!

Quoting the number of significant figures in the uncertainty

 As a general rule, where a rigorous method to determine the uncertainty has been used, the standard
uncertainty 𝑢(𝑧) should be quoted to two significant figures,
 for example: ± 0.032, ± 5.4, ± 72 x 103
 However, where a more casual approach (guessing) has been used, it is necessary to make a ‘common
sense’ decision and to quote the standard uncertainty 𝑢(𝑧) to one significant figure:
 for example: ± 0.04, ± 6, ± 8 x 104

Quoting the number of decimal places in the best approximation

Having quoted the uncertainty to two significant figures, count the number of decimal places to which this
corresponds, then write the best approximation to the same number of decimal places as the uncertainty.
For example:
𝑔 = (9.790 ± 0.052) ms −2 .

(has three decimal places here) and (two sig. figures and three decimal places here).

Quoting the result using fractional (%) uncertainty

It may be convenient on occasion to express a standard uncertainty 𝑢(𝑥) as a fraction of the best approximation
of the measurand 𝑥. Thus, we may sometimes refer to
𝑢(𝑥)
the fractional uncertainty given by , or
𝑥
𝑢(𝑥)
the percentage uncertainty given by × 100 %.
𝑥

For example, if the quoted result of a measurement is (3.60  0.32) units, then the fractional uncertainty is 0.32
/ 3.60 = 0.09 and the percentage uncertainty is 9%.
This form of quoting the uncertainty is commonly used by manufacturers of electronic components where the
uncertainty in the quoted value is called the “tolerance” and it is quoted as a percentage of the nominal value
of the component.
For example, if the value of a resistor is given as 220 Ω with a tolerance of 5%, it means that the resistance of
the resistor is (220 ± 11) Ω.

30
 Quoting experimental results as in many textbooks
There are other ways of quoting the results of experiments and a common method used in textbooks is, for
example, the quoting of the Universal Gas Constant R to be:
R = 8.31451(72) J K−1 mol−1.

The seven and two in brackets indicate that the last two digits ‘51’ have an uncertainty of ‘72’ associated with
them; in other words, the book value is:
R = (8.31451  0.00072) J K−1 mol−1.

Quoting the result with an expanded uncertainty

It sometimes happens that uncertainties are quoted to a greater coverage probability (i.e., with a greater
level of confidence) and this is done simply by multiplying the standard uncertainty by what is known as a
coverage factor, 𝑘.
When this is done the uncertainty is referred to as the expanded uncertainty, 𝑈, where 𝑈 = 𝑘𝑢.
The coverage factor, 𝑘, will depend on the coverage probability (level of confidence) with which the result is
to be quoted, as well as the way in which the standard uncertainty, 𝑢, was determined. The coverage factors
for the different combination are given below:
If a Gaussian pdf was used:
𝑘 = 1, and 𝑦 ± 𝑘𝑢 defines a 68% coverage probability;
𝑘 = 2, and 𝑦 ± 𝑘𝑢 defines a 95% coverage probability;
𝑘 = 3, and 𝑦 ± 𝑘𝑢 defines a 99% coverage probability.
If a rectangular pdf was used:
𝑘 = 1, and 𝑦 ± 𝑘𝑢 defines a 58% coverage probability;
𝑘 = 1.65, and 𝑦 ± 𝑘𝑢 defines a 95% coverage probability;
𝑘 = 1.73, and 𝑦 ± 𝑘𝑢 defines a 100% coverage probability.
If a triangular pdf was used:
𝑘 = 1, and 𝑦 ± 𝑘𝑢 defines a 65% coverage probability;
𝑘 = 1.81, and 𝑦 ± 𝑘𝑢 defines a 95% coverage probability;
𝑘 = 2.45, and 𝑦 ± 𝑘𝑢 defines a 100% coverage probability.
It is important to note that using a coverage factor does not change the result of the measurement; the
standard uncertainty remains unchanged. All that happens is that the result is presented in a different way.
The implication of this is that it is essential to report on the coverage probability (level of confidence) when
the result is quoted. So for example the value of 𝑔 on the previous page may be quoted as:
g = (9.79  0.16) m s−2 (99% coverage probability), where 𝑘 = 3.
Coverage probability should be kept consistent when comparing results modelled with different probability
distributions, i.e. say you directly compare results of measurements with Type A and Type B evaluations of
uncertainty. This could be achieved by increasing the coverage factor to 𝑘 = 2, which would calculate the
standard uncertainty for both evaluations with for a coverage probability of 95%.
Say that a measurement was made in order to determine the mass m of an object. The pdf of the result is
shown below in Figure 13(a) in which it may be seen that the best approximation of the mass 𝑚 is 83.50 g,
with a standard uncertainty of 0.12 g. This defines an interval 83.50  0.12 g within which we are 58%
confident that that the mass of the object will lie.

31
 Guide to Reporting and Measurement: January 2018

Figure 13: Probability density functions (pdfs) of a measurement corresponding to coverage factors
of 1 (a) and 2 (b).
The result in Figure 13(a) may be described as “𝑚 = (83.50  0.12) g, where the number following the symbol
 is the numerical value of the standard uncertainty 𝑢 = 0.12 g and defines an interval estimated to have a
coverage probability of 58 percent”.
The result in in Figure 13(b) may be described as “𝑚 = (83.50  0.19) g, where the number following the
symbol  is the numerical value of an expanded uncertainty 𝑈 = 𝑘 𝑢 with 𝑈 determined from a standard
uncertainty 𝑢 = 0.12 g and a coverage factor 𝑘 = 1.65, and defines an interval estimated to have a coverage
probability of 95 percent”.
We are 58 % confident that the value of the mass lies between 83.38 g and 83.62 g.
We are 95 % confident that the value of the mass lies between 83.31 g and 83.69 g.
We are 100 % confident that the value of the mass lies between 83.30 g and 83.70 g.
Note that by introducing an expanded uncertainty and coverage factor 𝑘 > 1, we are not affecting the result
of the measurement. We are only changing the way that we present the result.
Note in PHYLAB1 courses, it is rarely required to quote uncertainties to a greater coverage probability than
𝑘 = 1. We include this section here so you are aware of it for future study.

32
 D2 Comparing results
To say that two results are “close” or “nearly the same” is meaningless in the context of laboratory work.
The results of any two experiments can only be meaningfully compared if the intervals associated with each of
the results are known.
More specifically:
 If the intervals that represent the results of two measurements overlap, then we say these two results
“agree within experimental uncertainty.”
 If the intervals that represent the results of two measurements do not overlap, then we say these
results “do not agree within experimental uncertainty.”
For example, say three students measure the period 𝑇 of a pendulum and they each quote the result as
follows:
𝑇1 = (5.73 ± 0.41) s
𝑇2 = (5.62 ± 0.10) s
𝑇3 = (6.28 ± 0.25) s
The three measurements may be presented in the form of intervals on a number line:

5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.0 6.1 6.2 6.3 6.4 6.5 T (s)
| | | | | | | | | | | | | |

𝑇1 X

𝑇2 X

𝑇3 X

Figure 14: Comparison of results

 Note that the intervals associated with 𝑇1 and 𝑇2 overlap, and therefore “these two results agree
within experimental uncertainty”.
 The intervals associated with 𝑇1 and 𝑇3 also overlap so “these two results agree within experimental
uncertainty”.
 However, the interval associated with 𝑇3 does not overlap with the interval associated with 𝑇2 and
therefore “these two results do not agree within experimental uncertainty.”

33
 Guide to Reporting and Measurement: January 2018

D3 The uncertainty budget

An uncertainty budget is an evaluation (usually presented in the form of a table) of all the contributions of
uncertainty in a particular measurement (together with numerical estimates). These uncertainty components
are then used to calculate the combined standard uncertainty for the measurement result. An uncertainty
budget is a useful way to present how you dealt with all the uncertainties in a measurement.

Example 1: A measurement using an analogue voltmeter

Consider the situation where you are using an analogue voltmeter
to measure the voltage across the terminals of a battery. After
connecting the voltmeter across the terminals of the battery, what
you see on the voltmeter is shown alongside.
Say that you decide that the reading 𝑉 is 1.54 volts, and using a
triangular pdf, you determine the standard uncertainty 𝑢(𝑉read ) on this scale reading to be:
(1.57 − 1.51) V
𝑢(𝑉read ) = = 0.012 V.
2√6
It is usually not the case that you be 100 % confident that the instrument (in this case an analogue voltmeter)
gives a perfect reflection of the input (i.e. the voltage across the terminals of the battery). All analogue and
digital instruments have some internal calibration setting, referred to as the “rating” of the instrument. The
uncertainty associated with this rating is usually indicated by the manufacturers of the instrument and is often
quoted as a percentage.
Say that we are told that the meter we are using here has a percentage calibration uncertainty of 1%. Then
the uncertainty associated with the rating of the instrument is:
𝑢(𝑉rating ) = 0.01(1.54 V) = 0.015 V.
There may be other sources of uncertainty associated with the instrument, such as the influence of the
temperature of the surroundings, or the contact resistance in the probes. Let us say that these are negligible
in this case.
After you have determined all the possible sources of uncertainty and assigned a numerical value to each,
then you should draw up a table of the sort shown below, which is called an uncertainty budget. Each source
of uncertainty is listed together with its standard uncertainty. You should note that these standard
uncertainties can result from either Type A or Type B evaluations.
Table D3-1: Uncertainty budget for the single voltmeter reading, V.
Uncertainty component Standard uncertainty (V) Type of evaluation
Reading of the analogue 0.012 Type B
voltmeter display, 𝑢(𝑉read )
Rating of the analogue 0.015 Type B
voltmeter, 𝑢(𝑉rating )
Combined standard uncertainty: 𝑢(𝑉) = √(0.012 V)2 + (0.015 V)2 = 0.023 V.

The final result in this case may then be recorded as: “the best approximation of the voltage 𝑉 is 1.520 volts
with a combined standard uncertainty of 0.023 volts.”

34
 Example 2: A set of dispersed digital readings
Now let us presume that we are trying to measure the acceleration due to gravity 𝑔 by observing the period
𝑇 of a pendulum of length 𝑙. We determine that 𝑙 = 0.2619  0.0058 m (where 𝑢(𝑙) results from a Type B
evaluation of uncertainty with an analogue metre stick).
We then use a digital stopwatch to measure the period of the pendulum. You cause the pendulum to oscillate
and then record the time for 20 complete swings. You repeat this procedure 10 times and observe the data:
Table D3-2: Data recorded using a digital stopwatch.
Time for 20 swings, 𝑇20 (s) Period, 𝑇 = 𝑇20 /20 (s)
19.56 21.31 0.978 1.066
20.49 20.82 1.025 1.041
20.76 19.78 1.038 0.989
20.63 20.39 1.032 1.020
21.56 20.02 1.078 1.001

Since we observe a dispersion (scatter) in the readings, the best estimate of the period 𝑇 is given by the
arithmetic mean 𝑇̅, which is 1.027 s, and the standard uncertainty associated with the scatter is given by the
experimental standard deviation of the mean, which is 0.010 s. Then 𝑢(𝑇̅) = 0.010 s. This is a Type A
evaluation of uncertainty. Another source of scatter you might consider is the uncertainty in the gradient of a
linear graph (“the scatter in the (𝑥, 𝑦) data, 𝑢(𝑚)”).
Let us say that the manufacturers of the stopwatch report that it is accurate to 0.5%. Then the uncertainty
associated with the internal calibration or rating, 𝑢(𝑇rating ), will be:

𝑢(𝑇rating ) = (0.005)(1.027 s) = 0.0051 s.

To determine 𝑔, we use the formula:

𝑙 4𝜋 2 𝑙
𝑇 = 2𝜋√ → 𝑔= .
𝑔 𝑇2

This is our model equation for this measurement. The best approximation for 𝑔 is given by,
4𝜋 2 (0.2619 m)
𝑔= = 9.8028 ms −2 .
(1.027 s)2
The uncertainty budget for this measurement is shown below.
Table D3-3: Uncertainty budget for the measurement of gravitational acceleration, 𝒈.
Uncertainty component Standard uncertainty Type of evaluation
Scatter in the 𝑇 data, 𝑢(𝑇̅) 0.010 s Type A
Rating of the stopwatch, 𝑢(𝑇rating ) 0.0051 s Type B
Reading of 𝑙 from the analogue metre 0.0058 m Type B
stick, 𝑢(𝑙)
The combined uncertainty for 𝑇 is 𝑢(𝑇) = √(0.010 s)2 + (0.0051 s)2 = 0.011 s.
𝑢(𝑙) 2 𝑢(𝑇) 2
The standard uncertainty for 𝑔 is given by 𝑢(𝑔) = 𝑔√( ) +( )
𝑙 𝑇

0.0058 m 2 0.011 s 2
𝑢(𝑔) = (9.8028 ms −2 )√( ) + (2 ) = 0.302 ms −2 .
0.262 m 1.027 s

The final result may then be recorded as: “the best estimate of the acceleration due to gravity 𝑔 is 9.80 m s-2
with a standard uncertainty of 0.30 m s-2.”

35
 Guide to Reporting and Measurement: January 2018

SECTION E – FITTING A LINEAR FUNCTION TO THE DATA

E1 The process of fitting a function to a set of data

Sometimes you cannot directly measure the information you are seeking, so instead you measure other
values and determine the required information indirectly. This process often involves the method of ‘fitting’
some known mathematical function to a set of data.
The ‘fitting’ process is shown in Figure 15.
Measurement
Tabulate the data (readings).

Re-organise the data in accordance with the linearised function you

plan to use. See E2 Linearising equations.

Plot the graph to confirm that the expected relationship is indeed

linear.

Use a least squares fit to fit a straight line to the data, the equation of
which is 𝑦 = 𝑚𝑥 + 𝑐. See E3 Method of least squares.

Extract the information you are after from the coefficients m and c of
the fitted function.

Determine the uncertainties 𝑢(𝑚) and 𝑢(𝑐) by determining how well

the scattered data fits the mathematical function.
This is a TYPE A evaluation.

Quote the results of the measurement; 𝑚 ± 𝑢(𝑚) with units and 𝑐 ±

𝑢(𝑐) with units.

Calculation
Calculate the value of the measurand you are after and do the
necessary uncertainty calculations to determine the uncertainty in the
final result. See C1 Calculation of uncertaint.

Figure 15: Process for extracting information by fitting mathematical functions.

36
 E2 Linearising equations
When presented with a relationship between experimental variables that is not linear, the method of
‘linearising’ can be used extract information from recorded data.
Effectively, you rearrange the non-linear equation to take the form of a linear equation; 𝑦 = 𝑚𝑥 + 𝑐. After
plotting values on a graph for 𝑦 against 𝑥, you can use the gradient, 𝑚, and 𝑦-intercept, 𝑐 to calculate useful
values.
Example:

Consider an experiment to investigate the relationship between the position 𝑠 of an object over time 𝑡 while
that object is subject to a uniform (constant) acceleration 𝑎. This experiment is underpinned by the following
kinematic equation:

1 2
𝑠= 𝑎𝑡 + 𝑢𝑡 + 𝑠0 , … E2-1
2

where 𝑢 is the initial velocity and 𝑠0 is the initial position of the object.
If a set of data, (𝑡𝑖 , 𝑠𝑖 ), i = 1, 2, 3, …, n, were collected, then it would be difficult to determine the coefficients
𝑎, 𝑢 and 𝑠0 by plotting a graph of the equation E2-1 because the relationship between the position 𝑠 and the
time 𝑡 is a parabola of the form 𝑦 = 𝑎𝑥 2 + 𝑏𝑥 + 𝑐.
Equation E2-1 can be ‘linearised’ by considering the position relative to the initial position 𝑠0 and by dividing
throughout by 𝑡 ( given that 𝑡 ≠ 0 ):

𝑠 − 𝑠0 1 … E2-2
= 𝑎𝑡 + 𝑢.
𝑡 2
𝑠−𝑠0
In the rearranged equation, there is a linear relationship between and 𝑡, of the form, 𝑦 = 𝑚𝑥 + 𝑐:
𝑡
𝑠 − 𝑠0 1
= 𝑎 𝑡+𝑢
𝑡 2
𝑦 = 𝑚 𝑥+𝑐
𝑠𝑖 −𝑠0
Therefore plotting a graph with calculated values of (𝑡𝑖 , ) as the values (𝑥𝑖 , 𝑦𝑖 ) should yield a linear
𝑡𝑖
relationship.
You can now determine the acceleration, 𝑎, in equation E2-1 from the gradient of the graph and the initial
velocity, 𝑢, from the 𝑦-intercept.

37
 Guide to Reporting and Measurement: January 2018

E3 Method of least squares

An accurate method to find the slope 𝑚 and the intercept 𝑐 of the best fit line is to use the method of least
squares. This method also allows you to determine the uncertainty in the slope, 𝑢(𝑚), and the uncertainty in
the intercept 𝑢(𝑐).
Given a set of data pairs (𝑥𝑖 , 𝑦𝑖 ), i = 1, 2, 3, …, n, and that the data pairs represent a linear relationship, then
the slope 𝑚 and the intercept 𝑐 of the best fit line can be determined by the following equations:

𝑛 ∑ 𝑥𝑖 𝑦𝑖 − ∑ 𝑥𝑖 ∑ 𝑦𝑖
𝑚= … E3-1
𝑛 ∑ 𝑥𝑖 2 − (∑ 𝑥𝑖 )2

∑ 𝑥𝑖 2 ∑ 𝑦𝑖 − ∑ 𝑥𝑖 𝑦𝑖 ∑ 𝑥𝑖
𝑐= … E3-2
𝑛 ∑ 𝑥𝑖 2 − (∑ 𝑥𝑖 )2

Let 𝑑𝑖 = 𝑦𝑖 − (𝑚𝑥𝑖 + 𝑐), the distance between each data point and the line of best fit. The uncertainties
associated with m and c can be determined using:

∑ 𝑑𝑖 2 𝑛
𝑢(𝑚) = √ ( ), … E3-3
2 2
𝑛 ∑ 𝑥𝑖 − (∑ 𝑥𝑖 ) 𝑛 − 2

∑ 𝑑𝑖 2 ∑ 𝑥𝑖 2 𝑛
𝑢(𝑐) = √ ( ). … E3-4
2 2
𝑛(𝑛 ∑ 𝑥𝑖 − (∑ 𝑥𝑖 ) ) 𝑛 − 2

A computer is generally available to you in PHYLAB1 and the method of least squares can be employed using
the following programs:
 Microsoft Excel (by typing in the equations) or
 LinearFit (by entering the data).

If a computer is not available, use your calculator to draw up tables of data as suggested below and solve the
equations above. Some calculators will allow you to analyse data pairs.
𝑖 𝑥𝑖 𝑦𝑖 𝑥𝑖 2 𝑥𝑖 𝑦𝑖 𝑑𝑖 2

… … … … … …

… … … … … …

… … … … … …

∑ 𝑥𝑖 ∑ 𝑦𝑖 ∑ 𝑥𝑖 2 ∑ 𝑥𝑖 𝑦𝑖 ∑ 𝑑𝑖 2

38
 E4 Using EXCEL (2010 version)

Tables with Excel

Initiate Excel by clicking on the Excel shortcut on the computer’s desktop. Note the Excel worksheet is divided
into a series of columns (A-Z) and rows (1:99999). Each cell has an index corresponding to its column and row,
e.g. A2. If you want to get Excel to do a calculation, click on the cell where you want the answer to go, and
TYPE = followed by the required mathematical expression. For example, if the value in the particular cell is
obtained by multiplying the value in cell C3 by that in D3, TYPE = C3*D3 and press enter (or you can just TYPE
=, then click on cell C3, TYPE *, click on D3 and hit the enter key). In Excel, the * symbol is used to indicate
multiply, the / symbol divide, and the ^ symbol “raise to the power”. Use fractional powers for roots, e.g. to
the power of ½ is equivalent to taking the square root of a number.
To apply the same formula to a list of cells in a column; select the cell where you have typed the formula,
move the cursor over the bottom right hand corner until the + symbol appears, and then drag it down the
column. Excel copies the formula to all the cells and also adjusts the cell indices in each formula by
incrementing their row numbers in line with where you dragged the cursor. You can also copy horizontally,
which will increment the column letters in the cell indices. Try this out and click one of the copied cells to see
how the cell index has changed in the equation. If you want to fix the cell index to use the same value in all
formulae across different rows or columns, use a $ sign to fix a column and/or a row index (e.g. $E$5 included
in a formula is left unchanged when copied to a new cell).
Take care to display the correct number of significant figures in your data. Since Excel calculates the answer to
as many places as can physically fit in the cell, you often need to adjust the number of decimal places in
columns to reflect the correct number of significant figures. Increasing and decreasing the number of decimal

places can be achieved using the buttons marked . To outline the cells or add borders to your table

use the button marked . To add headings (and units) to columns, click on the cell and TYPE as you
would normally do. To produce superscripts and subscripts in the titles, select the text you want to raise or
lower, and use the right click to access the “Format Cells” menu. Then select “Superscript” or “Subscript”.
Further options can be accessed in the Excel menus or using the right-click menu.

Plotting with Excel

When you are ready for Excel to plot your data, you need to highlight the two columns of data you wish to
plot and use INSERT, SCATTER to call up the chart (plotting) procedure; ideally in the order 𝒙 then 𝒚. Select

the chart sub-type option which allows you to plot the points without lines. Under the CHART TOOLS set
of menus, you can change the Design (colour and style of data points), Layout (add titles, axes-titles, gridlines)
and the Format (chart size). Move your plot into the desired position by clicking just inside the borders and
dragging it to where you want it positioned. Other options can be accessed by selecting the chart with a right-
click of the mouse. You can change the origin and scale of the plot so that your graph covers most of the page.
To change the scale on the 𝑥-axis, click on the axis to select it. Next right-click to access the FORMAT AXIS
option and in the Axis Options sub-menu, adjust the minimum and maximum values as required. You can also
change the number of decimal places or to scientific exponent form under the Number sub-menu.

Best fit with Excel

To draw the line of best fit through your points, click somewhere inside your graph to select it, then click on
CHART TOOLS, LAYOUT, ADD TRENDLINE and select the LINEAR option (or alternatively select the data points
and access by the right-click mouse menu). If you need to know the slope and intercept of this straight line,
click on OPTIONS and check the DISPLAY EQUATION ON CHART box.

39
 Guide to Reporting and Measurement: January 2018

E5 Using LINEARFIT
LinearFit can be used to calculate the slope 𝑚 ± 𝑢(𝑚) and the intercept 𝑐 ± 𝑢(𝑐) of the linear line of best fit.
To start the program, click on the LinearFit shortcut on the desktop. The program will open and present the
page as shown in Figure 16. Type the 𝑥 and 𝑦 values into the appropriate columns, or alternatively import
your data in a .csv file (e.g. saved from Excel) in the format 𝑥, 𝑦, 𝑢(𝑥), 𝑢(𝑦) and LinearFit will automatically
calculate 𝑚 ± 𝑢(𝑚) and 𝑐 ± 𝑢(𝑐). When printing, don’t forget to add your student number to help you
identify your work at the printers.

Figure 16: Data input and fitting screen of LinearFit.

LinearFit with weightings

Tick the “Enable 𝑢(𝑥) weights” and the “Enable 𝑢(𝑦) weights” boxes. This will open two additional columns
for you to type in the weightings, i.e., the 𝑢(𝑥𝑖 ) and 𝑢(𝑦𝑖 ) values for the respective 𝑥𝑖 and 𝑦𝑖 readings, i = 1, 2,
3, …, n. When these weightings are included, the individual uncertainties are presented as intervals around
each data point; see Figure 17.

Figure 17: LinearFit with weightings enabled.

40
SECTION F – EXAMPLE OF FITTING A GAUSSIAN
Consider an example in which the measurand is the time taken for you to travel from wherever you live to the
university campus. So, for the next 40 trips to the university you use the stopwatch on your cell phone to
determine your travel time in minutes, i.e. you take a set of readings.
The data are tabulated as follows:
Table F-1: Time taken to get from home to campus (in minutes).

17.1 16.9 17.4 16.4 16.8 17.9 16.0 18.2

18.1 19.3 16.9 17.4 17.8 17.1 17.6 17.8
16.4 17.4 15.9 15.8 17.7 17.3 17.4 16.8
16.6 16.8 17.6 17.4 18.9 18.4 17.9 17.8
18.3 17.2 18.1 18.7 16.3 17.4 18.4 16.5

A simple calculation shows that the average time taken is 17.39 minutes.
We wish to present this data in a histogram, so the next step is to draw up a frequency table (or distribution
table). Note that in this example a bin width of 0.5 minutes was chosen. The choice of bin width is arbitrary
and is normally chosen to suit the data.
Table F-2: Frequency table for data in Table F-1.

Bin (minutes) Number of readings per bin

15.5 to 15.9 2
16.0 to 16.4 4
16.5 to 16.9 7
17.0 to 17.4 10
17.5 to 17.9 8
18.0 to 18.4 6
18.5 to 18.9 2
19.0 to 19.4 1
Total : 40
Next we plot a histogram showing how the data are distributed over the values of the readings taken by
plotting the number of readings per bin vs. the time interval of bins (see Figure 18). Finally, to show that the
general shape of the histogram suggests that the distribution of the data is Gaussian in shape, we
superimpose a trend line, shown as the light grey line in Figure 18.
It should be noted that as more and more readings are added to the set of data, and the bin width is
narrowed, the distribution of most data of this sort will tend towards the formation of a symmetrical Gaussian
distribution.

The Gaussian probability density function (pdf)

Having decided that the Gaussian probability density function (pdf) is a suitable probability density function
by which to model our knowledge of the measurand, we need to consider the implications. The probability
density function of a continuous random variable is a function that can be used to obtain the probability
that the random variable may take a specific value within some given interval.
When the probability density function (pdf) is normalised and portrayed graphically, the probability that
the random variable may take a specific value within a given interval is indicated by the area under the
graph subtended by that interval. So, if the given interval is all possible values, then the total area under
the pdf is “1” (unity), meaning that the probability that the specified value exists is also 1. As the interval is
made “narrower”, so the probability that the specific number is within that interval becomes less.

41
 Guide to Reporting and Measurement: January 2018

Figure 18: Histogram of travel times.

The important features of the Gaussian pdf are illustrated graphically in Figure 19.

Figure 19: Gaussian pdf showing standard deviation.

The equations used to calculate the standard deviation are given in B2 TYPE A evaluation of uncertainty, and
the standard deviation can also be found graphically by drawing a horizontal line at 0.61 of the maximum
height of the Gaussian, as shown in Figure 19.
In laboratory work, the use of the standard deviation gives a measure of the consistency of the measurement
process that was used in an experiment, and makes it possible to specify the level of confidence with which
the result of the measurement is quoted. In this case, the coverage probability (level of confidence) is 68%, as
explained in Figure 19.
However, exercise caution as the standard deviation and mean are only appropriate metrics for data that can
be modelled with a Gaussian distribution. You may encounter data that is better modelled with a different
distribution (e.g. radioactive decay is modelled with a Poisson distribution).

42