Appendix 2

Investigations, data handling and statistics

Science – how is it done?
Biology is a science. It relies on evidence in its attempts to explain and apply ideas about how the natural
world functions. People often talk about the ‘scientific method’ as if there is a single, unique method of
finding out, different from other methods of discovery. In fact there are many successful methods used by
scientists. This is clear when we look at the ways in which Charles Darwin, Louis Pasteur and Francis Crick
worked, for example. Nevertheless, all science includes some common features. These include:
• observing and measuring
• hypothesising and predicting
• designing and planning a sequence of observations, investigations or experiments
• carrying out these plans
• recording data
• interpreting results and drawing conclusions
• communicating.
Central to these is the concept of a hypothesis.
Sometime a hypothesis is formulated by examining data or from observations. It is equally likely to arise
from an inspired guess or a wild hunch. Whatever the origin of the hypothesis, valid and reliable evidence,
or data, is required to support it. This may come from well planned observations and experiments.

Hypothesis: A scientific hypothesis is a tentative explanation of a natural phenomenon capable of being tested by
observation or experiments.
 Extension
A hypothesis is never really proved but can be disproved.
It was Karl Popper, a philosopher of science who showed that a hypothesis is never really proved but can be
disproved at any time, by a single contrary observation. He maintained that all explanations of science are of this type.
That is, scientific knowledge is always only tentative. It is the best available explanation we can offer at any time. Any
explanation may be proved wrong or incomplete, and often is, sooner or later.

Obtaining and recording data

Types of data
In practicals and investigations, observations are recorded as data. There are different forms of data
(singular: datum), but they are all potentially useful. No one type of data that is relevant to your enquiry is
better than any other, providing they have been accurately made and recorded. The type of data we may
collect is either qualitative (as words) or quantitative (as numbers).
• Qualitative (or descriptive) observations – An example from classification studies is the colour or shape of
the parts of flowers. An example from behaviour studies is the feeding mechanism of honey bees visiting
flowers. Qualitative data may be recorded in written observations or notes, by photography or in drawings,
for example.
• Quantitative (or numerical) observations – Examples are the size (length, breadth or area) of the leaves of
a plant in shaded and exposed positions, and the pH values of soil samples in different positions.
In Table A2.1 this division is further classified.
Table A2.1 Categories of qualitative and quantitative data

Type of Type of data

variable

Qualitative

categoric nominal, i.e. values or observations belonging to it can be sorted according

to category, e.g. colour of flowers, gender

ordered ordinal, i.e. values that can be placed in an order or rank, the interval
between them may not be equal, e.g. opinion judgements – ‘completely
agree’, ‘mostly agree’, ‘mostly disagree’, ‘completely disagree’.

Quantitative – illustrated in Figure A2.2

continuous interval, that can have any value within a specific range and can be a whole
number, a fraction or a decimal. It can be counted, ordered and measured,
e.g. body mass, time taken for seeds to germinate after different treatments

discrete interval, that can have only a limited number of values which are usually
whole numbers, e.g. the number of seeds in a bean pod, the number of cells
in a haemocytometer grid
Extension
Handling very large and very small numbers
In science ‘powers of ten’ are used to avoid writing long strings of zeros when recording numbers. For example, the
age of the Earth, about 4 500 000 000 years, is written as 4.5 × 10 years. Similarly, a cyanobacterium, a
9

photosynthetic bacterium, may be about 0.000 003 6 metres in diameter, which is written as 3.6 × 10 m.
−6

This way of recording numbers is called scientific or standard notation. It is used to avoid the
errors that are easily made when writing down a large number of zeros. Also, when we need to
multiply numbers we can do so by adding powers. Similarly, to divide, the powers are subtracted.
Several of the powers of ten have prefixes which are frequently used in biology and are
represented by agreed symbols (Table A2.2).
Table A2.2 Powers of ten

× 10 kilo- or k × 10 deci- or d
3 −1

× 10 mega- or M × 10 centi- or c
6 −2

× 10 giga- or G × 10 milli- or m
9 −3

× 10 micro- or μ
−6

× 10 nano- or n
−9

These letters are written in front of the symbol

used to represent the quantity being measured,
such as kg, km, mm, nm.
Collecting and recording data
When you plan an investigation or experiment, think carefully about the data you expect to collect. This
allows you to prepare a recording sheet with spaces for the information (both qualitative and quantitative).
This will show how often you plan to record the data as well as exactly what will be recorded. An example
of such a recording sheet is illustrated in Figure A2.3.
Variables
In an investigation or experiment the specific conditions are called variables. Most of the variables are
deliberately kept the same, but others vary. In fact there are really three types of variable. We can
demonstrate them by referring to an example experiment, such as the investigation of the rate of an enzyme-
catalysed reaction, shown in Figure 3.8, page 62.
Remind yourself of this investigation, now.

Standardised (controlled) variables: the variables in an experiment or investigation that are kept the same so they
do not influence the measurement of the dependent variable.

The variables that are always kept the same in this investigation of the rate of reaction of catalase are the
conditions such as temperature, the volumes and concentrations of the reagents, and perhaps the pH (the
volume and type of buffer, if used, for example). Variables that are kept constant are called
the controlled or standardised variables.
Meanwhile, in Figure 3.8 you can see that measurements are made at 30 second intervals. So the times
that readings are taken are the variable that is manipulated by the experimenter, and is called
the independent variable. Note that it has been recorded in the table as a list before the experiment was
started.

Independent variable: the variable in an experiment or investigation that is manipulated or changed.

In the experiment, the amount of gas (oxygen) that has collected at 30 second intervals is the variable
accurately measured by the experimenter. It is called the dependent variable and is recorded in the table.
(Sometimes, two dependent variables may be measured. So, for example, in photosynthesis both oxygen
production and carbon dioxide intake may be measured.)
In a scientific investigation a control is used to ensure that any effects observed are due to changes in
values of the independent variable and not some unidentified variable. Thus a control is one value of the
independent variable. For example, in the investigation of the action of the enzyme catalase in bringing
about the breakdown of hydrogen peroxide, the control involves the use of boiled enzyme, since heat
denatures a protein and destroys its catalytic properties (Figure 3.10, page 64).

Dependent variable: the variable in an experiment or investigation that is measured.

Errors and the need for replicates

Errors and mistakes may occur in experiments and investigations, just as in any other human activity. You
cannot have complete confidence about the accuracy of every individual observation you make. For
example, an incorrect amount of reagent may accidentally be measured out and the mistake not noticed.
Alternatively, a stop watch may be misread or a reading from an instrument may be inaccurately taken on
some occasion. Meanwhile, you need confidence your results are accurate.
Consequently, the experiment should be repeated to provide another set of readings at the same time
intervals. If any of the measurements taken are wildly different you would be expected to further repeat the
experiment. Ideally, a minimum of three sets of measurements is required for minimum reliability.
Repetitions of an experiment are called replicates. So, in a class practical experiment carried out by you
and your peers, individual results may be pooled together as replicate readings. Replicate readings with
similar results give confidence in the conclusions drawn.

Reliability: reliable results are repeatable by the same student and reproducible by others.
Linear magnification and the actual sizes of
images and objects
Calculating the linear magnification of drawings and
photographs
Example
In a TEM of a red blood cell, the diameter of the image of the cell was 45 mm, and the actual size of the red
blood cell was 7.5 μm. What is the magnification of this TEM?
Magnification is the number of times larger an image is compared to its real size.

First, convert the observed size in mm to μm

45 × 1000 μm = 45 000 m
Then:

The ‘×’ symbol means ‘times’, so the magnification is recorded as × 6000.

To try
In a TEM of a mitochondrion, the length of the image of the organelle was 20 mm and the actual size was
5.0 μm.
What is the magnification of this TEM?

Calculating the actual size of an organism or structure

from a given scale bar
Example
In an interpretive drawing of Amoeba, the length of the image is 105 mm.
The scale bar is of length 22 mm and represents 0.1 mm.
What is the actual length of this Amoeba?
The scale bar represents 0.1 mm = 100 μm.
So the actual length of the Amoeba is:

This is approximately 480 μm.

To try
In a TEM of a chloroplast, the length of the image is 111 mm.
The scale bar is of length 10 mm and represents 8 μm.
What is the actual length of the chloroplast?
Determining the length of an appropriate scale bar for a
magnified image
Example
In an SEM, a mitochondrion of actual length 9 μm was magnified to a length of 45 mm.
Calculate how long a scale bar representing 5 μm would have to be on this SEM.
9 μm is equivalent to 45 mm.
45 × 5
So 5 μm is represented by

A scale bar representing 5 μm would be 25 mm in length.

To try
In a TEM of an animal cell, the nucleus of diameter 6 μm was magnified to a diameter of 135 mm.
Calculate how long a scale bar representing 1 μm would have to be on this TEM.

Calculating the size of specimens from magnified

drawings and photographs
Example
In a photomicrograph showing a human cheek cell, the image had been magnified ×800.
What is the actual length of that cell?
First: Measure the length of the image of the cell in the photomicrograph in millimetres, using a ruler.
It is 90 mm long at its maximum length.
Next: Convert this length to μm.

Then:

To try
In a photomicrograph of a green plant cell, the image had been magnified ×400. The image of the cell is 94
mm in length.
What is the actual size of the cell?
 Presenting data
Firstly, you must select from the data what is important. For example, it may be appropriate to round
numbers in order to avoid giving data to a greater level of accuracy than the measurements warrant. Then
you need to display the important data using a visual summary, making their importance clear. This might
involve graphs, bar charts, histograms, scatter graphs or pie charts, for example.

Graphs
Graphs show relationships or trends between two variables. It is conventional to plot the dependent variable
(the variable being measured) on the y (vertical) axis and the independent variable (the variable altered by
the experimenter) on thex (horizontal) axis. Both axes must be labelled and the units indicated. Points (a dot
within a circle or a cross, for example) must be plotted with a sharp pencil. The points should be joined by a
smooth curve only if you are confident that it indicates the likely points of intermediate readings. If you are
not confident of this, connect points with a straight line. If more than one line is plotted, then the plot points
and lines must be different and be clearly labelled. If the two plots have different vertical axes, then the scale
of the axes should be placed on either side of the graph.
The term graph applies to the whole graphic representation. The line on the graph showing a relationship,
whether straight or curved, is referred to as a curve. Examples of graphs are seen throughout this book.

Question
1 The following results were obtained in an investigation of the effects of pre-incubation of starch and
amylase solutions at different temperatures on the subsequent hydrolysis of the starch to sugar.

Plot a graph of these results, applying the rules and conventions detailed here.
Check your work against the model graph provided at the end of this Appendix.

Bar charts
Bar charts are used to show relationships between the independent variable (on the x axis) and the dependent
variable (on the y axis) when the independent is categoric and the dependent variable is continuous, such as
the range of tree species found in woods. There should be small gaps between the lines or bars used, which
should be of equal width, and typically presented in order of magnitude (Figure A2.4a).

Histograms
Histograms are useful to display continuous data. They show the variations in a sample of repeated
measurements. The x axis represents the variation in the repeated measurements. The y axis is the frequency
or number in each class. Normally the blocks are drawn touching. There should be an informative title
(see Figure A2.3).

Scatter graphs
We construct a scatter graph to investigate whether there may be a relationship between two variables. If one
variable increases, does the other also increase? This is a common situation in biology, for example, the
question of whether there is a correlation between blood pressure and heart rate, or smoking and heart
disease, or river flow rate and diversity of aquatic non-vertebrates. The first step in investigating a
correlation is to plot a scatter graph of one variable against another. The shape of the scatter graph indicates
the type of correlation. By ‘correlation’ we mean a mutual relation between two (or more) things, or an
interdependence of variable quantities. If both variables increase together then there is a positive
correlation; if one variable decreases when the other increases then there is a negative correlation.
 The closer the data points come to laying on a straight line, the closer the relationship; the more scattered
the data points the less close the relationship. If the scatter graph has random points then there is no
correlation (Figure A2.4b).
It is important to realise that a correlation between two variables does not necessarily mean that the
variable are causally linked. So, having applied a statistical test that indicates the possibility of a
correlation, we have to go on to investigate the mechanisms of the linkage, if there is one.

Pie charts
Pie charts are best used for showing relative proportions.
Statistical checks on data
Statistical tests should be used when you are not sure about the numerical relationships your data indicate.
The application of simple statistical tests is described below. Many calculators are programmed to carry out
statistical tests. Microcomputers will run spreadsheet programs with statistical tests programmed in.
Dedicated statistical software is available too.

The normal distribution

Data obtained from biological experiments may show a ‘normal distribution’. By this we mean that when
the frequency of particular classes of measurements (such as the number of humans at any particular height
is plotted against the classes of measurements (their different heights arranged in ascending order), a
symmetrical bell-shaped curve is obtained (Figure A2.5). With normal distributions, two issues may arise.
1 At what value do the readings cluster?
This can be expressed as:
• the average or arithmetic mean, calculated by dividing the sum of the individual values by the number of
values obtained

The formula for the arithmetic mean:

• the mode, the most frequent value in a set of values

• the median, the middle value in a set of values arranged in ascending order.
2 How spread out are the readings?
This can be expressed as the standard deviation(s) from the mean. This is a measure of the variation from
the mean of a set of values.
• A small standard deviation indicates that the data is clustered closely around the mean value.
• A large standard deviation indicates a wider spread around the mean.
Once obtained, the value may be applied to the normal distribution curve (Figure A2.6). Note that 68% of the
data occurs within one standard deviation of the mean and more than 95% of the data occurs within two
standard deviations of the mean. So, a small standard deviation indicates the observations (the values) differ
very little from the mean. Standard deviation is also a measure of reliability, so a small standard deviation
indicates reliability.

Standard deviation: the spread of a set of data from the mean of the sample is a measure of the variability of a
population from a sample. A small standard deviation indicates that the data is more reliable.

Calculating standard deviation

The standard deviation (s) is calculated in five steps.
1 First, calculate the mean, .
2 Calculate the deviation of each value from the mean: x − .
3 Square the deviations: (x − ) .
2

4 Add the squared deviations: ∑(x − ) .

2

5 Finally, divide by the number of values (n).

An example
An ecologist investigated the reproductive capacity of two species of a common grassland
flower Ranunculus acris and R. repens. The latter species spreads vegetatively via strong and persistent
underground stems. Would the use of vegetative reproduction result in fewer fruits from sexual reproduction
compared to the other species that reproduces only by fruit production?
Using comparable sized plants growing under similar conditions in the same soil, the numbers of fruits
formed in 100 flowers of each species were counted and recorded. The results are given in Figure A2.7 and
calculations of the standard deviations are shown in Figure A2.8. Note that you are not expected to know the
formula for calculating standard deviation. The purpose of presenting the steps in the calculation is to take
away the mystery of a calculation normally carried out by a scientific calculator or programmed spreadsheet.
Interpreting standard deviation
We have noted that a low value for the standard deviation indicates that the observations differ very little
from the mean and that a high value for the standard deviation indicates a wider spread around the mean.
Consequently, the standard deviation can be used to help to decide whether the differences between the
two related means, such as those shown in Figure A2.3 for example, are significant or not. If the standard
deviations are much larger than the difference between the means, then the differences in the means are
highly unlikely to be significant. On the other hand, when the standard deviations are much smaller than the
difference between the means, then the difference between the means is almost certainly significant.

Standard error: an estimate of the reliability of the mean of a population sample. A small standard error indicates that
the mean value is close to the actual mean of the population.
Standard error
The standard error (SM) represents how well the sample mean approximates the population mean. The
larger the sample, the smaller the standard error, and the closer the sample mean approximates the
population mean. The standard error is obtained by dividing the standard deviation, s, by the square root
of n, the sample size.

When graphs are presented showing mean values, error bars are added to each value plotted to demonstrate
the deviation of the sample from the true population mean. Error bars (±SM) extend above and below the
points plotted on a graph to show this variability.
The t-test
Statistical tests using standard deviation or standard error typically compare large, randomly selected
representative samples of normally distributed data. In practice it is often the case that data can only be
obtained from quite small samples. The t-test may be applied to sample sizes of more than 5 and less than 30
taken from normally distributed data. It provides a way of measuring the overlap between two sets of data –
a large value of t indicates little overlap and makes it highly likely there is a significant difference between
the two data sets. The following example illustrates the method of the t-test, however, you should note that
you are not expected to calculate values of t.

Extension
The null hypothesis
Statistical tests are hypothesis-testing statistics. They test a mathematical statement called the null hypothesis.
Where we are comparing data the null hypothesis states that there is no difference between the sets of data. When
we are looking for an association it states that there is no association.
The outcome of a statistical test is a probability that the null hypothesis is true. A probability
(known as the P-value) varies from 0 (impossible) to 1 (certain). Since the P-values are small, they
are given as a percentage (0 to 100%) to avoid possible confusion with small numbers. The lower
the probability, the less likely it is that the null hypothesis is true.
The t-test makes comparison between means of data to test for significant differences between
the samples. For the t-test the null hypothesis is ‘There is no difference between the means’. By
convention in biology, if the probability is greater than 0.05 (5%) then the null hypothesis is
accepted. However, if the probability is 0.05 or less (P < 0.05), then the null hypothesis is rejected.
This implies the event is predicted to happen by chance less than once in twenty times. So the
difference is judged to be significant.

Applying the t-test

An ecologist was investigating woodland microhabitats, contrasting the communities in a shaded position
with those in full sunlight. One of the plants was ivy (Hedra helix), but relatively few occurred at the
locations under investigation. The issue arose: were the leaves in the shade actually larger than those in the
sunlight?
Leaf widths were measured but, because the size of the leaves varied with the position on the plant, only
the fourth leaf from each stem tip was measured. The results from the plants available are shown in Table
A2.3.
Table A2.3 Widths of leaves of Hedera helix in sun and shade

Size-class widths of leaves from widths of leaves from

/ mm plants in sun (a) plants in shade (b)

20–24 24

25–29 26, 26 26

30–34 30, 31, 31, 32, 32, 33 33, 34

35–39 37, 38 35, 35, 36, 36, 36, 37

40–44 43 41, 42

45–49 45
 The steps of the t-test
1 The null hypothesis assumes the difference under investigation has arisen by chance. That is, there is no
difference in width between leaves from plants growing in sun and shade. The role of this statistical test is to
determine whether to accept or reject the null hypothesis. If it is rejected in this case, we can have
confidence that the difference in the leaf sizes of the two samples is statistically significant.
2 Next, check that the data are normally distributed. This is done by arranging the data for the two samples as
in Table A2.3. (and plotting a histogram, if necessary).
3 You are not expected to calculate values of t. This is a statistic which, if required can be found by using a
scientific or statistics calculator or by means of a spreadsheet incorporating formulae.
Actually, the formula for the t-test for unmatched samples (data sets a versus b) is:

where:
a = the mean of data set a
b = the mean of data set b
sa = the standard deviation of data set a, squared
2

sb = the standard deviation of data set b, squared

2

na = the number of data items in set a

nb = the number of data items in set b
4 Once a value of t has been calculated (the value of t in this case is 2.10) we need to consult a table of critical
values for the t-test, once we have determined the degrees of freedom (d.f.) for the two samples, using the
formula:

In this case, there are 11 + 11 = 22 degrees of freedom.

5 A table of critical values for the t-test is given in Figure A2.10.
Look down the column of significance levels (p) at the 0.05 (5%) level until you reach the line corresponding
to 22 degrees of freedom. You will see that in this case, p = 2.08.
6 Since the calculated value of t (2.10) exceeds this critical value (2.08) at the 0.05 (5%) level of significance, it
indicates that there is a lower than 0.05 (5%) probability that the difference between the two means is solely
due to chance. Therefore, we can reject the null hypothesis, and conclude the difference between the two
samples is significant. (For the experimenter, the significance of this statistic suggests there is a reason for
the difference in the means. This can be further investigated and perhaps a fresh hypothesis proposed).
 The chi-squared test
χ is the Greek letter chi. We have previously introduced and applied the chi-squared (χ ) test to discover
2

whether observed numerical results differ from the expected numerical result (Table 16.5, page 365). It tests
for ‘goodness of fit’ between an observed distribution and a theoretical one. It allowed us to test whether the
observed results obtained from the dihybrid cross between Drosophila of normal flies (wild type) with flies
homozygous for vestigial wing and ebony body differ significantly from the expected outcome.
Turn back to pages 364–5 and refresh your memory of the formula for the chi-squared statistic and how it
is applied. Note that in the chi-squared test, the null hypothesis is ‘Observed frequencies equal expected
frequencies’. The chi-squared test is also applied in ecology for looking at the differences in distribution of
organisms in different habitats.

In summary
We have now introduced the four tests that you may need to be able to apply. The formulae for all four are
given in Figure A2.11.
You do not need to memorise the formulae or the meaning of their symbols. However, you may need to
use them in practical work:
• to calculate a standard deviation
• to put error bars on graphs
• to test for a significant difference between the means of two small samples
• to perform a chi-squared test on suitable data.
To do this you will have access to the formulae, the meaning of the symbols, a t-table and a chi-squared
table. Rather than carry out all the steps of a test in an examination, you may be given partly completed
calculations to finish. Consequently, it is helpful to be fully acquainted with the use of an approved
electronic calculator and have used it to become familiar with each of the four tests.
Answer to question 1