Pillay - 2016
Probability Notes
Recall that = 1 ( 1 )
If you see E(Y2) anywhere as a given value remember
!
() = ( ! )
If the word, differs, is used, do the modulus of the variable in the P
brackets, e.g. X.
Remember that if ~ , = ! + ! + ! , Y consists of r
independent observations of X, therfore, ~ (, )
For questions asking for the ( + ) or the expectational equivalent,
find the variance or expectation of each. For Y= ! + ! + ! ! , find
variance just for X and then do () = (), because n
independent observations of X.
If its something with only 1-5 possible values, do a table like this, e.g. for
! !
! = + ! ,
! !!
2 1
( = )
= 3 , ~(3, 0.2)
>5
This means that he can get a maximum of 2 hits within the 5 throws he has
available, hence = 5 .
~ 5, 0.2 , 2 = 0.942 = ( 5)
Estimators
An estimator is not unbiased if
! ! is an biased estimator for population variance
!
!! = !
!
!! = !
! 1
! = ! ! + ! ! + ! ! !
! = ! ! (1)
!
!
= = ! !
!
+ != ! 2
Subbing 1 into 2
! ! !
!! = ! ! = ! ! =
! !
( 1)
!! = ! =
Therefore, because ! , ! is a biased estimator of ! , to convert it
! ! !
to an unbiased estimator
Note the following
! ( ! )!
!! = !
=
( ! )! = ! !
The most efficient estimator is the one with minimum variance, so do the
!!!
derivative and prove it is the minimum using !! ! .
If asked to create an estimator for k or something like that, do E(X), find it in
terms of k, then make k the subject.
If possible, set ! and factorise it out, markschemes award marks
for this, and it is easier later when you do the derivative of (), where V
is the estimator under investigation.
Ardon Pillay - 2016
Confidence Intervals
When we have a k% CI, we are saying that there is a k% chance that the
variable being investigated is in that interval
Out of 100 samples, k of them will have our variable in them
It is taken from a sample of size n
Derivation
1.96 < < 1.96 = 0.95
1.96 < < 1.96 = 0.95
1.96 < < 1.96 = 0.95
1.96 < < + 1.96 = 0.95
1.96 < < + 1.96 = 0.95
0.95 = 1.96 , + 1.96
Mean ()
! !
o General CI for = ! ! , + ! ! for k%. This is only if we
can obtain an unbiased estimate for the variance (a.k.a we are not
explicitly given the variance). The symbol is only used if the
sample comes from a normal distribution. Otherwise, we define
the standard deviation as !!! , the square root of the unbiased
estimator of population variance.
o ! is found by working backwards from the CI. E.g. if it is 95%.
This means that < < = 0.95, hence < =
!!!.!"
!
= 0.025. Then apply inversenorm to find the value of a = !
o , , ! !
o , if we have a T distribution to work with,
o Margin of error = ! !!
o Width of CI = 2 ! !! or
o If given an interval like 32.5, 33.5 , use width to find either or .
Can do sum of upper and lower point, this will be equal to 2,
hence can find sample mean
Ardon Pillay - 2016
Proportion (p)
o To find the CI of a proportion, we use ! or ! , and then apply the
following CI formula.
! !!! ! !!!
o % = ! !
, + ! !
, where is the
sample proportion
o
o = ,
~(, (1 )
! !!!
o % = ! !
Continuous Random Variables
1. Basics
!""#$ !"#"!
a. !"#$% !"#"$
= 1
!""#$ !"#"$
b. = !"#$% !"#"$
!""#$ !"#"$ !""#$ !"#"$ !
c. = !"#$% !"#"$ ! !"#$% !"#"$
!"!"#$
d. !"#$% !"#"$
= 0.50
!"! !"#$"%&'(" !
e. !"#$% !"#"$
= !""
f. Interquartile range: 75 25
g. Mode is the x value that begets the highest vertical axis value
(differentiate)
h. For CDF
i. If the c.r.v has the pdf
0 0
3
= ,0 1
5
3
3 !, 1 < < 3
20
For 0 1
!! ! ! !
! !
= = ! 0 1 , where t is a dummy
! !
variable
For 1 3
!
3 3
1 + 3 !
5 ! 20
!
3 1 3 3 ! 8 3 !
+ 3 ! = + = + 1
5 ! 20 5 20 20 20
Ardon Pillay - 2016
0 0
3
0 1
= =
3 !
+ 1 1 < < 3
20
1 > 3
0.75 , 0.75 ,
i. VI: Bringing in Y. We must change X into ( < ), then change
( < ) into ( < ), where a is an expression in terms of y.
This gives us a CDF for Y, which can be differentiated with respect
to y to give a PDF for Y. See below
!! !
Question: = = = !"
, 1 < < 3. Find the CDF of X. If =
! !
!
, > !
For 1 < < 3
! !
3 ! 3 !
= = =
! 26 !
3(26) ()
1 1 1
= = 1 = =1
1 1 1 27 1 1
=1 =1 !
= !
, 1 3
26 26 26 26
New bounds
3 3 27 1
> =1 =1 + ! =
4 4 26 3
26 4
PGFs the sub-topic to rule them all
1. The Basics
a. = ! = ! =
b. 1 = 1 = 1! = = = = 1
c. ! 1 = = 1 !!! = = =
!
! ! !
d. = 1 + 1 1
!! !
e. ( = ), use the Maclauren Series: ( = ) = !!
f. !
g.
2. Finding the Probability Density Function (PDF)
a. First, we expand ! = and look for a pattern within the
coefficients of t. We then can use that pattern to produce a general
term for the coefficients, which as we know, will equal = .
Ardon Pillay - 2016
3. Tricks
a. The coefficient of ! is simply ( = )
b. 2 binomials, if they have the same p, produce a binomial when
combined
c. No matter what, if we combine 2 poissons, itll be a Poisson
d. ! = !
e. If asked to prove the relationship between 2 variables, and PGFs
have been used in the question prior to it, use PGFs to prove
f. See notebook for proof for why + = ! (), really
easy to make careless mistakes so keep your head and be
confident.
Expectation and Variance
1. Poisson - = , =
2. Binomial - = , = (1 )
3. Normal - = , = !
! !
4. Geometric - = !, = !!
! !"
5. Negative Binomial - = !, = !!