In a box plot outliers are

1.5 interquartile ranges above the upper quartile

or In a histogram, total area under bars is

PROPORTIONAL to total frequency.
1.5 interquartile ranges below the lower quartile

Always refer to outliers or extreme values but not

anomalies.

The three properties of a normal distribution

PMCC is not affected by linear coding.
are
So multiplying every value of x or y by the same
1. The curve is bell shaped
(or indeed, adding, subtracting, dividing) does not
2. The curve is symmetrical about the mean
affect the PMCC
3. mean = mode = median

Positive skew mode < median < mean

Events are independent if they do not affect each LOOK AT THE TAIL!!
others outcomes Median closer to lower quartile than higher
quartile
P(A∩B) = P(A) x P(B) Higher frequencies at lower values
Or you could say

P(A|B)= P(A) or
P(B|A) = P(B)

Negative skew mean < median < mode

LOOK AT THE TAIL!! When finding the lower quartile work out ¼ n
Median closer to upper quartile than lower quartile
Higher frequencies at higher values If this is a decimal use the first value after (e.g.
6.5 you would use 7th value)

If this isn’t a decimal use the average of this

value and the one after (e.g. 7 you would
average 7th and 8th value)
 Modelling: 7 steps! RDPECER
A statistical model is Step 1: Recognise a real world problem
Step 2: Devise a statistical model
A statistical process to describe or Step 3: model used to make Predictions
Step 4: Experimental data is collected
make predictions about the expected Step 5: Comparisons made against devised model
bahaviour of a real-world problem Step 6: Evaluation: Statistical concepts used to test
how well model describes real-world problem
Step 7: Refine model

Why use statistical models?

If you are given E(X) then
E(aX) = aE(X)
Used to simplify or represent a real world problem
E(aX + b) = aE(X) + b
Cheaper or quicker or easier or more easily modified
Example E(X) = 3
To improve understanding of the real world problem
Then E(5X)= 15
Used to predict outcomes from a real world problem
E(7X+4)=25

If you are given Var(X) then

Var(aX) = a2Var(X)
Var(aX + b) = a2Var(X) In a stem and leaf diagram, put all leaves in
order and remember the key!
Example Var (X)= 4
Var (2X)= 22 x 4 = 16 A title with what the diagram is showing is also
Var (2X - 3)= 22 x 4 = 16 helpful.
Var (3 - 2X)= (-2)2 x 4 = 16

Events are mutually exclusive if

P( A | B ) = P( A ∩ B ) P(A∩B) = 0
P ( B)
A and B cannot happen together
P(something given something) = P(both)/ P(given)

Use a histogram if the data is continuous, and if Also can be written as

the data is grouped unevenly.
P ( A ∩ B ) = P ( A) + P ( B ) − P ( A ∪ B )
If a distribution is skewed then it has extreme values
Put your answer in context
Median/ Interquartile range better than mean/
standard deviation because they are not affected by Positive correlation (not enough)
outliers. However the mean takes into account all
values! Positive linear correlation (better)

Normal distribution not suitable for modeling skewed Strong positive linear correlation; as
distributions as it is symmetrical. temperature increased so did number of ice
creams sold (best)

Mean =

When talking about skew in a box plot look only at

the box!! ∑ x or ∑ xf
n ∑f

Standard deviation is

or
A statistical experiment is a
∑x 2
∑x f
2

− mean
2
− mean2 process used for collecting
n ∑f
data to provide evidence for or
LEARN THIS against a hypothesis
Mean of squares – square of mean

It is not suitable to extrapolate (estimate

An event is using a line of best fit outside the range of data
collected) as a linear relationship may not
remain valid.
one of the possible outcomes Interpolation is more accurate but is
of an experiment dependent on how strong the correlation is
(look at the PMCC)

A model will never be able to

A residual is the vertical distance between a
cater for all the eventualities of a point on a scattergraph and a line of best fit!
real life problem
 Standardise by taking away mean and dividing
For normal distributions draw a diagram.
by standard deviation.

If X ~ N(4, 9), find P(X<12). First we "standardise". The When finding the upper quartile work out ¾ n
mean is 4 and σ is 3 (the square root of 9).
If this is a decimal use the first value after (e.g.
6.5 you would use 7th value)
(subtracting 4 and dividing
each side of the inequality by 3) If this isn’t a decimal use the average of this
= P(Z < 8/3)= Φ (8/3) = Φ (2.67) = 0.9962 value and the one after (e.g. 7 you would
average 7th and 8th value)

Capital F just means the probability to the left

of and inclusive! It is the cumulative probability.

F(0.6) = P(X<0.6)

Always remember that

When finding the median work out ½ n
F(LAST POSSIBLE VALUE) = 1
If this is a decimal use the first value after (e.g. 6.5
you would use 7th value) e.g. if F(x) = for x = 0,1,2,3 then a
( x + 2) 2
If this isn’t a decimal use the average of this value
and the one after (e.g. 7 you would average 7th and 25
8th value) table would be

Do the same for quartiles and percentiles x 0 1 2 3

F(x) 4/25 9/25 16/25 1

Then a probability distribution would be

x 0 1 2 3
p 4/25 5/25= 7/25 9/25
1/5

A MEASURE OF SKEWNESS IS

3(MEAN – MEDIAN)/STANDARD DEVIATION

All lines of best fit must go through the mean
POSITIVE- POSITIVE SKEW

NEGATIVE- NEGATIVE SKEW

 The Normal distribution is continuous so you
If possible draw a venn diagram to help you!
can’t do P(X = a) as this would be ineligible.
REMEMBER THE BOX.
For example P(X < a) = P(X < a)

A discrete uniform distribution is a probability

distribution with the same probability of all outcomes
occurring

e.g rolling a fair die

The assumption that all probabilities are equal is
Useful in theory – allows problems to be modelled If in a venn diagram, you are given that
something has occurred, look only at the
However it is not necessarily true in practice appropriate circle.

To refine the model you would carry out an experiment to

establish probabilities

Be prepared to interpret the gradient and y intercept

of a regression line IN CONTEXT
The explanatory variable is the variable you
e.g. if we are talking start with (the one you control)

x axis – years employed by company The response variable is the variable which
y axis –salary occurs as a result of the behaviour of the
explanatory variable.
Then gradient would be salary increase per year
y intercept would be starting salary
x -1 0 1 2 3
p 0.3 0.2 0.2 0.2 0.1

Mean=E(X)=
Only use a regression line is the PMCC shows
strong correlation , as this shows the data can be ∑ xp = (− 1× 0.3) + (0 × 0.2) + (1× 0.2) + (2 × 0.2) + (3 × 0.1)
modeled by a straight line. = -0.3+0.2+0.4+0.3=0.6
Variance = E(X2) – (E(X))2=
> +0.7
< -0.7 ∑x 2
p − mean2
=
( − 1) 2 × 0.3 + (0 2 × 0.2) + (12 × 0.2) + (2 2 × 0.2) + (3 2 × 0.1) − 0.6 2
= 0.3+0.2+0.8+0.9-0.36=1.84
 Some features of a box plot are
To work out E(X2) just do the first bit of the
variance formula 1. Allows comparisons
2. Shows outliers
3. Indicates spread/range/IQR
∑x 2
p
4. Shows max/min/median/quartiles
5. Shows skewness

Don’t round until the end; never use a rounded When using formulae, show your substitutions
value in calculations carefully

This will lose you marks THIS WILL GET YOU MARKS FOR WORKING

Stem and leaf can also be used to show skewness

On a box plot
High frequency at lower values POSITIVE SKEW
Label axes and scale (ie values)
High frequency at high values NEGATIVE SKEW

If you are asked to work out a complex

A GOOD WAY TO SUMMARISE 2 SETS OF probability, THINK!!
DATA IN ONE DIAGRAM IS A BACK TO BACK
STEM AND LEAF AS IT ALLOWS List outcomes or sketch a tree diagram.
COMPARISON. Can the situation happen in more than one
way?

Coding does affect mean and standard deviation

Multiplying all values by 3 would increase standard

deviation by x 3 and mean x 3

Dividing all values by 3 would divide mean and

standard deviation by 3

Adding or subtracting 3 would add or subtract 3 to

the mean but would not affect the standard
deviation