S1 Cheat Sheet

Using Your Calculator

You can handily use your calculator’s STAT mode to calculate Getting Started Once in calculation mode, if you want some statistical expression
the mean and variance of a frequency table, find 𝑟 (PMCC) in your calculation, e.g. 𝑥̅ or 𝑟 or Σ𝑥, press SHIFT  1 (for STAT).
Press MODE  STAT. “1-Var” means ‘one variable’. Use
and the 𝑎 and 𝑏 for the line of best fit, along with a number
“𝑎 + 𝑏𝑥” for two variables (e.g. regression and Var: Contains 𝑥̅ , 𝑛 (same as Σ𝑓), 𝜎𝑥
of summations such as Σ𝑥, Σ𝑥𝑦, Σ𝑥 2 , …
correlation). Note that frequency is not a variable.
Sum: Σ𝑥, Σ𝑥𝑦, Σ𝑥 2 , Σ𝑦 2
Setting Up:
Entering Data
Reg(ression): 𝑎, 𝑏, 𝑟 (using for regression and correlation
Press SHIFT  MODE for setup. Press down, select STAT
Press = after each number is entered. Use the arrow keys chapters).
then turn frequency mode On (your calculator will preserve
to navigate around the table. Press AC once finished to go
this setting even when turned off).
to calculation mode.
Exam Note: Suppose you need to calculate 𝑆𝑥𝑥 . It is fine to use your calculator to get summations like Σ𝑥 and so on. But you must still show the substitution of the numbers into the relevant
formula. The calculator is incredibly handy for checking the value of 𝜎, 𝑎, 𝑏, 𝑟, 𝑥̅ , but should be used to check, not to do the question for you.
See http://www.drfrostmaths.com/resource.php?id=11591 for a Virtual Silver Calculator giving more detail on the buttons and modes available.

Chapter Usual types of questions Tips What can go ugly

1 – Mathematical  “What are the stages of a  Stages are (memorise!)  Not much provided you
Models mathematical model?” 1. Recognition of a real-world problem. memorise what’s on the
 “Write down reasons for 2. Statistical model is devised. left! Such exam questions
using mathematical 3. Model used to make predictions. are rare but students have
models.” 4. Experimental data is collected. been thrown because they
5. Comparisons are made against the devised model. hadn’t expected to be
6. Statistical concepts are used to test how well the model describes the real-world tested on Chapter 1, or
problem. their teacher has skipped
7. Model is refined. the chapter.
 Reasons for using statistical models (memorise!)
1. Simplify a real-world problem.
2. To improve understanding/describe/analyse a real-world problem.
3. Quicker and cheaper than using real thing.
4. Predict possible future outcomes.
5. Refine model/change parameters possible.

2/3 – Measures of  Calculating mean and  Calculating 𝑥̅ is two marks. The method mark is for the division, NOT showing your calculation of So much can go wrong! From
Location and standard deviation from a midpoints times frequency. Writing out the calculation in full takes too much time and prone to experience, students:
Dispersion grouped frequency table. Σ𝑥
calculator-punching errors. Use STAT mode to get Σ𝑥, then just show as your calculation (i.e.  misremember the standard
𝑛
 Calculating quartiles and with the Σ𝑥 and 𝑛 subbed in), followed by the final answer. Do not round your mean or give as a deviation formula (either
the median using fraction! (the latter which can be penalised) forgetting to square the
interpolation. Note that for grouped data Σ𝑥 on your STAT mode is Σ𝑓𝑥 and 𝑛 gives Σ𝑓. mean, or forgetting to
square root the variance to

www.drfrostmaths.com 1
  Stating the effects of  Mmemonic for variance: “The mean of the squares minus the square of the mean (msmsm)” get the standard
coding on quartiles/median Σ𝑥 2 deviation).
1. Ungrouped data: 𝜎 2 = − 𝑥̅ 2
/mean/standard deviation Σ𝑓𝑥 2
𝑛  don’t observe whether
/interquartile range. 2. Grouped data: 𝜎 2 = − 𝑥̅ 2 there are ‘gaps’ in the
Σ𝑓
 Be able to combine means. Don’t forget to square root for variance! And use your exact value for 𝑥̅ . Note that Σ𝑥 2 on your intervals and thus are
calculator is effectively Σ𝑓𝑥 2 when frequencies are given. penalised heavily for linear
Standard deviation can roughly be thought of as “the average distance of values from the mean” interpolation.
(e.g. if you just had 3 and 7 as values, the standard deviation IS 2 because they’re both 2 away  introduce rounding errors
from the mean of 5). Thus do a common-sense check that your value looks sensible. because they don’t use
 Coding: Note that you’d always calculate statistics like the mean on the coded data, not the their exact mean in the
original data. The idea is then you work out what this stat would be on the original data. standard deviation
e.g. If coding is 𝑦 = 𝑥 − 2, where original variable is 𝑥 and new coded one 𝑦, we’d calculate say calculation.
𝑦̅, then add 2 to get the original mean 𝑥̅ . Make sure therefore that you check whether the  think that Σ𝑥 2 means
question wants you to work forwards or backwards. (Σ𝑥)2 .
1. Mean/median/quartiles: Affected by all of ×,÷, +, −  round the item number for
2. Standard deviation/range/interquartile range: Only affected by ×,÷ linear interpolation. If 𝑛 =
3. Variance: Any scale factors will be squared. e.g. If 𝑦 = 3𝑥 + 2, 𝜎𝑦2 = 9𝜎𝑥2 40 and you’re interpolating
 To work out the item number required for say the lower quartile: to find the LQ, use the 10th
𝑛 𝑛 item: don’t do ANY
1. For listed data: Find (similarly for median, or 0.34𝑛 for 34th percentile, etc.)
4 2
adjustment! Similarly, if
If a decimal, round up. If whole, use halfway between this item and one after.
𝑛 𝑛 = 35, use 8.75, not 9!
2. For grouped data: Find . DO NOT ROUND OR ADJUST. Then use linear interpolation.
4
 Linear Interpolation
Be vigilant of gaps in the intervals vs no gaps. If gaps, just expand the intervals first so fill the
space. So 3 − 5 would become 2.5 – 5.5. Note this increases the class interval by 1.

12
Suppose we want median. = 6 so use 6th item (this is grouped data so don’t do anything to the
2
6). Construct a suitable line. I put units on the interval values to avoid confusion between
frequencies and values of the variable (in this case weight).

3 3
Since 6 by observation is of the way along the interval frequency-wise, we want to go of the
4 4
way along the class interval. Thus:
3
𝑄2 = 5.5 + ( × 3) = 7.75𝑘𝑔
4
𝑛1 ̅̅̅
𝑥1̅+𝑛2 ̅̅̅
𝑥2̅
 Combined means: 𝑥̅ = (General principle is to add the sums of the values and divide by
𝑛1 +𝑛2
the combined number of things there are)

www.drfrostmaths.com 2
4 – Continuous  Constructing a box plot  There’s 3 ways you can find skew:  Not showing your outlier
Random Variables (probably with outliers). 1. For histograms/probability distributions, just observe the shape. If the ‘tail’ is in the boundary calculations for
 Construct a stem-and-leaf positive direction, then positive skew. box plots.
diagram (or back-to-back 2. Using quartiles: 𝑄3 − 𝑄2 > 𝑄2 − 𝑄1 means right box in box plot is larger, so positive  Referring to things you
stem-an-leaf). skew. haven’t calculated (e.g.
 Calculate interquartile 3. If positive skew then 𝑚𝑒𝑎𝑛 > 𝑚𝑒𝑑𝑖𝑎𝑛. I remember as the mean determining the skew; quartiles) when justifying
range/median from a stem- e.g. large salaries in the positive tail drag mean up but not median, so positive skew. what the skew of your data
and-leaf diagram. It is vital you use whatever data is available from the previous parts of the question. If you’ve just is.
 Find widths/heights of bars worked out the quartiles but don’t know the mean, then it would be stupid to give your  Not considering gaps (if
in a histogram. justification of the skew involving mean and median. present) for histograms.
 Find the frequency of a  For box plots, definition of outlier boundary is always given, but generally:  Just being careless when
interval using a histogram 𝐿𝑄 − 1.5 × 𝐼𝑄𝑅 and 𝑈𝑄 + 1.5 × 𝐼𝑄𝑅 finding areas within a
(e.g. number of runners You must explicitly show the calculation you used to calculate your boundaries, and there’s histogram (e.g. misreading
with a time under 10s). marks associated with this. Outliers are indicated using × symbol. It’s possible to have multiple scales).
 Use histogram to complete outliers at each end (or none).  Getting positive and
frequency table and vice-  There’s two possibilities for the end points of the whiskers when there’s an outlier on that end, negative skew the wrong
versa. and mark schemes accept both: either use the outlier boundary itself, or the smallest/greatest way round.
 Use frequency table value which is not an outlier (I prefer the latter). But be consistent in which you use.
formed from histogram to  For histogram questions involving finding width and height of bars, just find the scale factor
make subsequent Chp2/3 between class width and actual width, and between frequency density and actual height.
calculations, e.g. mean. e.g. If a bar with class width 3 and frequency density 10 has ‘drawn’ width 6cm and height 5cm,
 Calculate and interpret then the scale factors are 2 and 0.5 respectively. Then if some other bar has class width 5 and
skew. frequency density 8, then its width and height are 10cm and 4cm.
 Remember that in histograms, area is not necessarily equal to frequency, they are proportional.

Key is to use a known area with known frequency to find the scale factor 𝑘. This scale factor can
then be used to convert other areas in the histogram into frequencies. (See my Chp4 slides)
 As per Chps2/3, you must check whether the intervals have ‘gaps’, and adjust appropriately
𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦
when calculating frequency density. 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 =
𝑐𝑙𝑎𝑠𝑠 𝑤𝑖𝑑𝑡ℎ
 When asked to find the mean, median, quartiles or variance of a histogram, first use the
histogram to generate a grouped frequency table. Then use this table as you usually would to
calculate these statistics
 “Explain why a histogram is appropriate for this data”  “Data is continuous.” (No credit given
for talking about different possible class widths)
 Possible wordy questions:
“Explain whether the mean or median would be more suitable for this data.”
“The median because the data is skewed”. The question wouldn’t be asked if the data wasn’t
skewed. The logic is that extreme values affect the mean but not the median, and thus the
median may be more suitable.

www.drfrostmaths.com 3
5 – Probability  Find probabilities by  One neglected type of question, usually found at the end of Discrete Random Variable exam  Nonsensical notation, e.g.
enumerating outcomes. questions, is to be able to enumerate different matching outcomes so we can find the total 𝑃(0.3), when you mean
 Use Venn diagrams to find probability. that the probability is 0.3.
frequencies. Suppose we had the following distribution:  Using 𝑃(𝐴 ∩ 𝐵) =
 Use Venn diagrams to find 𝑃(𝐴)𝑃(𝐵) when you
unknown probabilities. weren’t told the events
 Use the laws of Then if 𝑋 is defined as the product of two spins of this spinner, we might be interested in finding were independent, or
probabilities to calculate 𝑃(𝑋 > 2). As per GCSE, we could list the possible outcomes, find the probability of each, then similarly 𝑃(𝐴 ∪ 𝐵) =
probabilities, e.g. involving add them together. You can be clever however in how you combine possibilities: 𝑃(𝐴) + 𝑃(𝐵) when you
conditional probabilities, weren’t told they were
independent and mutually 𝑆1 𝑆2 𝑃(𝑆1 × 22 ) mutually exclusive.
exclusive events, etc. 1 ≥3 0.1 × 0.7 = 0.07  Just generally getting stuck
 Use tree diagrams to 2 ≥2 0.2 × 0.9 = 0.18 and not knowing how to
determine probabilities. ≥3 𝑎𝑛𝑦 0.7 × 1 = 0.7 proceed to calculate a
 Be able to construct a Venn probability given other
Diagram that takes into Thus 𝑃(𝑋 > 2) = 0.07 + 0.18 + 0.7 = 0.95. See Ex1 in my slides: ones. As per the advice, try
account the relationship http://www.drfrostmaths.com/resource.php?id=11650 drawing a Venn Diagram.
between events. Were you told any
 Venn Diagrams with frequencies are fairly self-explanatory. Just be on the lookout for the word relationships between the
‘given that’ when asked to find a probability. e.g. If “given that they are cyclist”, then your events? (i.e. mutually
probability is out of the cyclist set/circle, rather than out of everyone. exclusive or independent)
 Ensure you can correctly identify regions in a Venn Diagram involving intersections, unions and  Accidentally doing
𝑃(𝐴∩𝐵)
complements, e.g. 𝐴 ∩ 𝐵′. Recognise that (𝐴 ∪ 𝐵)′ is the same as 𝐴′ ∩ 𝐵′. 𝑃(𝐴|𝐵) =
𝑃(𝐴)
 When asked to find probabilities given other probabilities, Venn Diagrams are particularly a good  Writing your conditional
strategy if probabilities involve intersections and complements, e.g. if you know 𝑃(𝐴) and probability wrong in the
𝑃(𝐴 ∩ 𝐵) and need to find 𝑃(𝐴 ∩ 𝐵′), then via a Venn Diagram it would be much easier to see first place, e.g. if “Given
that 𝑃(𝐴 ∩ 𝐵′ ) = 𝑃(𝐴) − 𝑃(𝐴 ∩ 𝐵). that A, find probability of
B”, then 𝑃(𝐴|𝐵) would be
All Laws of Probability: wrong. Remember the “|”
 If 𝑨 and 𝑩 are mutually exclusive (i.e. “can’t happen at the same time”) then: symbol means “given that”,
𝑃(𝐴 ∩ 𝐵) = 0 so check by reading the
𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) symbols out to yourself.
 If 𝑨 and 𝑩 are independent (i.e. “one event does not influence the other”) then:
𝑃(𝐴 ∩ 𝐵) = 𝑃(𝐴)𝑃(𝐵)
𝑃(𝐴|𝐵) = 𝑃(𝐴)
 In general:
𝑃(𝐴∩𝐵)
𝑃(𝐴|𝐵) = (I remember this as ‘the probability of the intersection divided by the
𝑃(𝐵)
probability of the thing you’re conditioning on’)
𝑃(𝐴 ∪ 𝐵) = 𝑃(𝐴) + 𝑃(𝐵) − 𝑃(𝐴 ∩ 𝐵) (the Addition Rule)

www.drfrostmaths.com 4
  There are certain reactionary things you should write out immediately after reading the initial
blurb of a probability question, before even starting part (a) of the question:
1. If you see that “𝐴 and 𝐵 are independent”, immediately write out 𝑃(𝐴 ∩ 𝐵) =
𝑃(𝐴)𝑃(𝐵), as you might forget you can use this otherwise.
2. If you see a conditional probability given, write out the conditional probability formula
involving those events.
 Always be on the lookout for the words ‘given that’. If so, you’ll have a conditional probability
and hence need to use the appropriate formula. Tends to come up at end of question.
 Do you have a mixture of 𝑃(𝐴 ∩ 𝐵), 𝑃(𝐴 ∪ 𝐵), 𝑃(𝐴) and 𝑃(𝐵) in the problem? Write out the
Addition Rule and sub things in. There’s harder 4 mark questions that involve use of this.
 You are sometimes asked to construct a Venn Diagram which encapsulates the relationship
between events. Remember that if events are mutually exclusive, the circles must be separate.
But if you’re not told they’re mutually exclusive, you must assume they can overlap.
Independence does not affect the Venn Diagram.

 Suppose we had a Tree Diagram where we had event 𝐴 followed by event 𝐵 followed by 𝐶.
1. 𝑃(𝐶)? As per GCSE, find all paths which match and each probabilities for each path.
𝑃(𝐶) = 𝑃(𝐴 ∩ 𝐵 ∩ 𝐶) + 𝑃(𝐴 ∩ 𝐵′ ∩ 𝐶) + 𝑃(𝐴′ ∩ 𝐵 ∩ 𝐶) + 𝑃(𝐴′ ∩ 𝐵′ ∩ 𝐶)
2. 𝑃(𝐵)? Note we need not consider 𝐶 as it happened after 𝐵.
𝑃(𝐵) = 𝑃(𝐴 ∩ 𝐵) + 𝑃(𝐴′ ∩ 𝐵)
3. 𝑃(𝐶|𝐵)? Harder tree questions ask you to calculate some conditional probability. Just
use the conditional probability formula as normal.

Students tend to find probability questions the hardest. The key is just practising the four different
types of questions. See my slides http://www.drfrostmaths.com/resource.php?id=11650
6 – Correlation  Calculate 𝑆𝑥𝑥 , 𝑆𝑥𝑦 , 𝑆𝑦𝑦  Ensure you know what your 𝑥 variable is and what your 𝑦 variable is, if the question uses  Forgetting to square root
 Calculate the PMCC, 𝑟 different letters. The first row in any table given is always the 𝑥. denominator in formula for
 Interpret a correlation  “Interpret the correlation coefficient”. ‘Interpret’ means a non-statistical wordy interpretation is 𝑟.
coefficient. required, e.g. “As the altitude increases, the temperature increases”.  Stating the correlation (e.g.
 State the effect of coding  Generally >0.7 and <-0.7 is considered strong correlation. “negative correlation”)
on PMCC.  “Is their claim justified?”. Generally two marks. e.g. “Claim justified (1 mark) as the value of 𝑟 is when you were asked to
 Comment on whether a close to 1 (1 mark)”. interpret it.
given claim about  All linear codings (adding/multiplying/dividing/subtracting) have no effect on PMCC.
correlation is justified.  𝑆𝑥𝑥 is affected by coding in the same way as variance is. So if we used 𝑡 = 3𝑥 + 1, then
𝑆𝑡𝑡 = 9 × 𝑆𝑥𝑥
 Use the STAT mode to check your value of 𝑟 if the original data is given. You obviously still need
to show your calculations for 𝑆𝑥𝑥 and so on.

www.drfrostmaths.com 5
7 – Regression  Given data or summarised  𝑦 = 𝑎 + 𝑏𝑥 then 𝑏 =
𝑆𝑥𝑦
and 𝑎 = 𝑦̅ − 𝑏𝑥̅ . Use your exact value of 𝑏 in your calculation for 𝑎;  Mixing up your 𝑥 and 𝑦
𝑆𝑥𝑥
data, find the line of best variables (particularly when
avoid rounding errors! If the question uses variable letters other than 𝑥 and 𝑦, match up the
fit 𝑦 = 𝑎 + 𝑏𝑥 they use different variable
variables first so you don’t get 𝑆𝑥𝑥 and 𝑆𝑦𝑦 mixed up. The first row is always 𝑥.
 Comment whether a letters, and hence
 Remember that your calculator’s STAT mode can check your values of 𝑎 and 𝑏 if the original data 𝑆𝑥𝑦
prediction made using your incorrectly using 𝑏 =
is provided. You’d be a donut not to check. 𝑆𝑦𝑦
𝑦 = 𝑎 + 𝑏𝑥 formula is
reliable.  Worded questions: instead of 𝑏 =
𝑆𝑥𝑦
𝑆𝑥𝑥
1. “Explain why this diagram would support the fitting of a regression line of 𝑦 onto 𝑥.”
 Find the equation of the  Forgetting to write out the
 “The points are close to an implied straight line of best fit.”
line of best fit using the equation of the regression
2. “Interpret the gradient 𝑏”
original non-coded line after working out 𝑎
 “As [the altitude] increases by 1, [the temperature] increases/decreases by ___” (This
variables. and 𝑏.
needs to match the context)
 Be able to ‘interpret’ the  Accidentally mixing up 𝑎
3. “Interpret the y-intercept 𝑎”
values of 𝑎 and 𝑏. and 𝑏. Remember that 𝑦 =
 “[The temperature] is 𝑎 when [the altitude] is 0”
 Justify which is the 𝑎 + 𝑏𝑥. You may be used
4. “Which is the explanatory variable. Explain your answer.”
explanatory variable. to 𝑦 = 𝑚𝑥 + 𝑐 from GCSE,
 “[The altitude] is the explanatory variable as [the altitude] influences [the
 Explain why the data where the 𝑥 term generally
temperature]”
justifies fitting a regression comes first. In stats, the 𝑥
 Reliability questions: (After using your equation 𝑦 = 𝑎 + 𝑏𝑥 to estimate a 𝑦 value)
line. terms comes second.
“Comment on the reliability of your estimate.” One of:
1. “Reliable (1 mark) as value inside the range of the data (1 mark)”
2. “Unreliable (1 mark) as value outside the range of the data (1 mark)”
3. “Reliable (1 mark) as value just outside the range of the data (1 mark)”
 Coding is simple; simply sub in the expressions and rearrange.
e.g. If 𝑦 = 3 + 2𝑥 and you had coding 𝑥 = 3𝑎 + 4 and 𝑦 = 𝑏 − 1, then:
𝑏 − 1 = 3 + 2(3𝑎 + 4)
𝑏 = 12 + 6𝑎
8 – Discrete  Find a constant 𝑘 using a  Note the difference between a probability function and a probability distribution:  When calculating 𝐸(𝑋 2 ) for
Random Variables probability 0.1𝑥, 𝑥 = 1,2,3,4 use in 𝑉𝑎𝑟(𝑋) formula,
distribution/function. Example probability function: 𝑝(𝑥) = { remember that the
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
 Find a constant 𝑘 using a Equivalent probability distribution: outcomes only are
cumulative distribution squared, not the
function. probabilities.
 Turn a probability  Common error:
distribution into a  Note the difference between shorthand 𝑝(𝑥) (with lowercase 𝑝) and longhand 𝑃(𝑋 = 𝑥) (with 𝑉𝑎𝑟(3𝑋) = 3𝑉𝑎𝑟(𝑋)
cumulative distribution, capital 𝑃).
and vice versa.  Recall that 𝐹(𝑥) = 𝑃(𝑋 ≤ 𝑥) i.e. the running total of the probability. If given a decimal value
 Calculate the expected (and outcomes are integers), then round down, i.e. 𝐹(3.7) = 𝐹(3).
value or variance of a  When asked to find a missing constant (say 𝑘) in a probability function, first find the probability
discrete random variable. of each outcome (i.e. find the probability distribution), then use fact that probabilities sum to 1.
 Find two missing  When similarly asked the same of cumulative distribution function, use fact that 𝐹(… ), for the
probabilities (using last outcome is 1.

www.drfrostmaths.com 6
 simultaneous equations)  If given an inequality within a 𝑃(… ), solve the inequality first. e.g. 𝑃(2𝑋 + 1 < 7) = 𝑃(𝑋 < 3)
when the expected value is  𝐸(𝑋) = Σ𝑥 𝑝(𝑥) and 𝑉𝑎𝑟(𝑋) = 𝐸(𝑋 2 ) − 𝐸(𝑋)2 (the “msmsm” mnemonic still applies here)
given.  If you have two missing probabilities in a probability distribution, and are given the expected
 Calculate 𝐸(𝑋 2 ). value, use: (a) Fact that probabilities add up to 1 and (b) Equation for expected value. This gives
 Find the probability of an you two simultaneous equations in terms of these variables.
inequality involving a  Note that rearranging the expected value equation gives 𝐸(𝑋 2 ) = 𝑉𝑎𝑟(𝑌) + 𝐸(𝑌)2 . On two
discrete random variable. occasions in the past, exam questions have asked for 𝐸(𝑋 2 ).
 Coding: Same rules apply as with other chapters.
1. For 𝐸(… ), tip is to rub out E then write it back round 𝑋, e.g. 𝐸(3𝑋 + 1) = 3𝐸(𝑋) + 1
2. For 𝑉𝑎𝑟(… ), remember scale factors are squared, and + and – have no effect.
1 1
e.g. 𝑉𝑎𝑟(3𝑋 + 1) = 9𝑉𝑎𝑟(𝑋) and 𝑉𝑎𝑟 ( 𝑋 − 1) = 𝑉𝑎𝑟(𝑋)
2 4
 Don’t forget that discrete random variable questions can have a second half involving probability
questions more like Chapter 5. See notes on this section.
9 – Normal  Find the probability of a For most of the examples below we will use the example of IQ, which by definition has the  Getting the sign of your z
Distribution value being more/less than distribution 𝑋~𝑁(100,152 ) value wrong.
a value. Some core stuff first:  Being sloppy with
 Find the probability of a  𝑧 represents the number of standard deviations above the mean. So an IQ of 𝑥 = 130 rearrangement and doing
10
value being in a two-ended corresponds to 𝑧 = 2 given it’s two standard deviations above the mean of 100. Similarly 𝑥 = 85 say = 20 and doing 𝜎 =
𝜎
range, i.e. 𝑃(𝑎 < 𝑋 < 𝑏) 𝑋−𝜇 20
gives 𝑧 = −1. It’s therefore not difficult to see therefore why 𝑍 =
 Find the value that gave 𝜎 10
rise to a probability.  The z-table allows you to find the probability up to a given z-value. The first table requires the  As with probability, just
 Find the value in a double- inequality to be <, the z-value to be positive and the probability to be at least 0.5. writing nonsensible
ended range, e.g. 𝑎 such  If in 𝑃(𝑍 < 𝑧) you either have to change the direction of the inequality or the sign of z, you do 1 notation, e.g. using
that 𝑃(𝜇 < 𝑋 < 𝜇 + 𝑎) = minus. If you change both, the “1 minuses” cancel each other out. e.g. 𝑃(𝑍 < 0.6) when 0.6 is the
0.3 or 𝑃(𝜇 − 𝑎 < 𝑋 < 𝜇 + 𝑃(𝑍 < 2) → 0.9772 𝑃(𝑍 < −2) = 1 − 𝑃(𝑍 < 2) = 0.0228 probability.
𝑎) = 0.4 𝑃(𝑍 < −2) = 1 − 𝑃(𝑍 < 2) = 0.0228 𝑃(𝑍 > −2) = 𝑃(𝑍 < 2) = 0.9772  Forgetting to standardise.
 Find conditional  The same applies if using the z-table backwards. If using the first table, you require the  Not using the second z-
probabilities involving the probability to be at least 0.5 and the inequality to be <. table when you are able to
normal distribution. 𝑃(𝑍 < 𝑧) = 0.58 → 0.2 do so (this costs you at
 Find a missing 𝜇 and/or 𝜎. 𝑃(𝑍 < 𝑧) = 0.42 → 𝑃(𝑍 < −𝑧) = 0.58 → −𝑧 = 0.2 → 𝑧 = −0.2 least a mark).
𝑃(𝑍 > 𝑧) = 0.58 → 𝑃(𝑍 < −𝑧) = 0.58 → −𝑧 = 0.2 → 𝑧 = −0.2
(note in the last example above we had to change both the direction and the sign to ensure 0.58
stays above 0.5)
𝑃(𝑍 > 𝑧) = 0.42 → 𝑃(𝑍 < 𝑧) = 0.58 → 0.2
 The approach above is the ‘super safe’ but slightly cumbersome approach. Easier is to think
whether we’re below the mean or above the mean. If you know you’re below (e.g. you know
you’re in the bottom 30%), the z value is negative, otherwise positive.
 You MUST use the value from the second z-table if you have a ‘nice’ probability. e.g.
𝑃(𝑍 < 𝑧) = 0.2 → 𝑧 = −0.8416 (as explained above, we’re in the bottom 20% so we know
we’re in the bottom half)

www.drfrostmaths.com 7
  Questions involving z-table being used forwards.
“Find the probability that someone has an IQ above 130.”
𝑃(𝑋 > 130) (Express question probabilistically)
130−100 𝑋−𝜇
= 𝑃 (𝑍 > ) = 𝑃(𝑍 > 2) (Standardise using 𝑍 = )
15 𝜎
= 1 − 𝑃(𝑍 < 2) (Manipulate so we can use table)
= 0.0228 (Use table)

“Find the probability that someone has an IQ between 85 and 130.”

𝑃(85 < 𝑋 < 130)
= 𝑃(𝑋 < 130) − 𝑃(𝑋 < 85) (in general 𝑃(𝑎 < 𝑋 < 𝑏) = 𝑃(𝑋 < 𝑏) − 𝑃(𝑋 < 𝑎) )
= 𝑃(𝑍 < 2) − 𝑃(𝑍 < −1)
= 𝑃(𝑍 < 2) − (1 − 𝑃(𝑍 < 1)) …

 Questions involving z-table being used backwards.

“70% of people have an IQ above you. What is your IQ?”
Quick way: Slow but safer way:
𝑃(𝑋 > 𝑥) = 0.7 𝑃(𝑋 > 𝑥) = 0.7
𝑥−100
= −0.5244 (as we’re in the bottom half) 𝑃(𝑍 > 𝑧) = 0.7 (standardise)
15
𝑥 = 92.134 𝑃(𝑍 < −𝑧) = 0.7
𝑥−100
𝑧 = −0.5244 = −0.5244
15
𝑥 = 92.134

(I will do the ‘quick’ way from now on)

“Find the value of 𝑎 such that 𝑃(100 < 𝑋 < 𝑎) = 0.3”

Key is to convert two-ended range into just one-ended. With a quick sketch it’s clear that
𝑃(𝑋 < 𝑎) = 0.8
(Alternatively we could have done 𝑃(100 < 𝑋 < 𝑎) = 𝑃(𝑋 < 𝑎) − 𝑃(𝑋 < 100) = 0.3
𝑃(𝑋 < 𝑎) − 0.5 = 0.3 → 𝑃(𝑋 < 𝑎) = 0.8
We can then proceed in the usual way.

“Find the value of 𝑎 such that 𝑃(𝜇 − 𝑎 < 𝑋 < 𝜇 + 𝑎) = 0.3

With a quick sketch we can see we have an area symmetrical about the mean. If 0.3 is in the
middle then we have 0.35 at each tail. Thus 𝑃(𝑋 < 𝜇 + 𝑎) = 0.65
𝜇+𝑎−𝜇 𝑎
= = 0.39 → 𝑎 = 5.85
15 15

www.drfrostmaths.com 8
  Questions involving a missing 𝝁 and/or 𝝈
“If 𝑋~𝑁(70, 𝜎 2 ) and 𝑃(𝑋 < 80) = 0.6, find the value of 𝜎.”
Method is pretty much exactly as before:
80 − 70
= 0.2533
𝜎
10
𝜎= = 39.48
0.2533

“If 𝑋~𝑁(𝜇, 𝜎 2 ) and 𝑃(𝑋 > 20) = 0.3 and 𝑃(𝑋 < 10) = 0.25, find 𝜇 and 𝜎.”
20 − 𝜇 10 − 𝜇
= 0.5244 = −0.67
𝜎 𝜎
20 − 𝜇 = 0.5244𝜎 10 − 𝜇 = −0.67𝜎
Then subtracting: (being very careful about your negatives)
10 = 1.1944𝜎
𝜎 = 8.3724
𝜇 = 20 − 0.5244 × 8.3724 = 15.6095

 Questions involving conditional probabilities.

𝑋~𝑁(100,152 ). Given that someone’s IQ is over 85, find the probability their IQ is over 115.
𝑃(𝑋 > 115 ∩ 𝑋 > 85) 𝑃(𝑋 > 115)
𝑃(𝑋 > 115 |𝑋 > 85) = =
𝑃(𝑋 > 85) 𝑃(𝑋 > 85)
𝑃(𝑍 > 1) 0.1587
= = = 0.1886
𝑃(𝑍 > −1) 0.8413
Note that 𝑋 > 115 ∩ 𝑋 > 85 is equivalent to 𝑋 > 115, as 𝑋 > 85 is already implied by 𝑋 > 115.
I call these “redundant events”. They come up quite frequently in S1 exams.

www.drfrostmaths.com 9