PROBABILITY THEORY

AND MATHEMATICAL
STATISTICS
As.prof. Oksana Pavlenko

2020.
The formula of total probability and LECTURE 2
Bayes’ formula

O.Pavlenko. Probability theory. 2020.

 LECTURE 2
The formula of total probability

H1, H2, ..., Hn - a partition of a sample space.

P( H1 ) + P( H 2 ) + ... + P( H n ) = 1

A = H1 A + H2 A + ... + Hn A
n
P( A) = P( H1 A) + P( H 2 A) + ... + P( H n A) =  P( H i A)
i =1

P( H i A) = P( H i ) P( A / H i )

n
P( A) =  P( H i ) P( A / H i ) – the formula of total probability
i =1
Remark. The total probability can be interpreted as weighted mean value of
conditional probabilities.

O.Pavlenko. Probability theory. 2020.

Example 1 There are two boxes with computer chips. The first
box contains 10 valid chips and 2 defective; the second box
contains 5 valid chips and 2 defective. 3 random chips are moved
from the first to the second box. Compute the probability that
among 2 chips taken at random from the second box only one is
defective.

3
𝐶10 C81  C 21
𝐻1 : vvv 𝑃(𝐻1 ) = 3 ≈ 0.55 P( A / H1 ) = 2
 0.36,
𝐶12 C10
2
𝐶10 ⋅ 𝐶21 C71  C31
𝐻2 : vvd 𝑃(𝐻2 ) = 3 ≈ 0.41, P( A / H 2 ) = 2
 0.47,
𝐶12 C10
1
𝐶10 ⋅ 𝐶22 C61  C 41
𝐻3 : vdd 𝑃(𝐻3 ) = ≈ 0.04, P( A / H 3 ) = 2
= 0.53,
3
𝐶12 C10

3
P( A) =  P( H i ) P( A / H i )  0.55  0.36 + 0.41 0.47 + 0.04  0.53  0.41
i =1

O.Pavlenko. Probability theory. 2020.

Example 2 A computer assembling company receives 25% of
parts from supplier X, 40% of parts from supplier Y, and the
remaining 35% of parts from supplier Z. Two percent of parts
supplied by X, five percent of parts supplied by Y, and ten
percent supplied by Z are defective. If some chosen at random
part is not defective, what is the probability that this part was
received from supplier X?
P( H1 / A) − ?
P( H 1 ) = 0.25, P( A / H 1 ) = 0.98,
P( H 2 ) = 0.4, P( A / H 2 ) = 0.95,
P( H 3 ) = 0.35 P( A / H 3 ) = 0.9

O.Pavlenko. Probability theory. 2020.

P( H i / A) − ? P( H i A) = P( AHi )

P( H i ) P( A / H i ) = P( A) P( H i / A)

P( H i ) P( A / H i )
P( H i / A) = – Bayes’ formula
P( A)

Thomas Bayes

«An Essay towards solving a Problem in the Doctrine of Chances» 1763.g.

O.Pavlenko. Probability theory. 2020.

Example 2 A computer assembling company receives 25% of
parts from supplier X, 40% of parts from supplier Y, and the
remaining 35% of parts from supplier Z. Two percent of parts
supplied by X, five percent of parts supplied by Y, and ten
percent supplied by Z are defective. If some chosen at random
part is not defective, what is the probability that this part was
received from supplier X?
P( H1 / A) − ?
P( H 1 ) = 0.25, P( A / H 1 ) = 0.98,
P( H 2 ) = 0.4, P( A / H 2 ) = 0.95,
P( H 3 ) = 0.35 P( A / H 3 ) = 0.9
3
P( A) =  P( H i ) P( A / H i ) = 0.25  0.98 + 0.4  0.95 + 0.35  0.9 = 0.94
i =1

P( H1 ) P( A / H1 ) 0.25  0.98
P( H1 / A) = =  0.26
P( A) 0.94

O.Pavlenko. Probability theory. 2020.

Example. Real estate agency employs four agents: A, B, C and D. Agent A
serves 10% of clients, agent B – 20%, but agents C and D serve equal
number of clients. 60% of A’s clients, 35% of B’s, 45% of C’s and 40% of
D’s clients sign a deal.
Take a client who signed a deal. What is the probability that he was served by
agent B? Find this probability for agent A, agent C, agent D.
H1 – a client was served by agent A, P( H1 ) = 0.1, P( A / H1 ) = 0.60,
H2 – a client was served by agent B. P( H 2 ) = 0.2, P( A / H 2 ) = 0.35,
H3 – a client was served by agent C, P( H 3 ) = 0.35, P( A / H 3 ) = 0.45,
H4 – a client was served by agent D.
P( H 4 ) = 0.35. P( A / H 4 ) = 0.40.
A – a client signed a deal.
P ( H 2 / A) − ? 0.1 0.6
P( H1 / A) =  0.14
n 0.4275
P( A) =  P( H j ) P( A / H j ) = 0.1 0.6 + 0.2  0.35 + 0.2  0.35
j =1 P( H 2 / A) =  0.16
0.4275
+ 0.35  0.45 + 0.35  0.4 = 0.4275 0.35  0.45 =1
P( H 3 / A) =  0.37
0.4275
P( H j ) P( A / H j ) 0.35  0.4
P( H j / A) = P( H 4 / A) =  0.33
P( A) 0.4275

O.Pavlenko. Probability theory. 2020.

Bernoulli scheme of independent trials
n is the number of independent trials,
P(A)=p is the probability of some event A in each trial (experiment),
m is the number of successes,
q=1-p is the probability of non-success (or failure).

Bernoulli formula 𝑃𝑛 (𝑚) = 𝐶𝑛𝑚 𝑝𝑚 𝑞 𝑛−𝑚

Example A fair die is thrown 7 times. What is the probability

that the number of points is 3 at least 2 times? 1 5
p= q = 1− p =
6 6
𝑷𝟕 (𝒎 ≥ 𝟐) = 𝑷𝟕 𝟐 + 𝑷𝟕 𝟑 + 𝑷𝟕 𝟒 + 𝑷𝟕 𝟓 + 𝑷𝟕 𝟔 + 𝑷𝟕 𝟕

𝑃7 (𝑚 ≥ 2) = 1 − 𝑃(𝑚 < 2) = 1 − 𝑃7 0 − 𝑃7 1 ≈0.33

1 6
1 5
0 7
1 5
P7 (1) = C71        0.39 P7 (0) = C70        0.28
6 6 6 6

O.Pavlenko. Probability theory. 2020.

Bernoulli scheme of independent trials

Example It is known that a marksman doesn’t hit a target in three

trials in a row with the probability 0.216. Compute the
probability that he will have 5 hits in the series of 7 shots.

P3 (0) = 0.216 𝑃7 5 −?

P3 (0) = C30  p 0  (1 − p)3 = (1 − p)3 = 0.216

p = 0.4

𝑃7 (5) = 𝐶75 ⋅ 0.4 5

⋅ 0. 62 ≈ 0.08

O.Pavlenko. Probability theory. 2020.

Bernoulli scheme of independent trials

The Mode is the number of successes with the largest probability

np − q  Mo  np + p

Example n = 7,
p = 0.4
7  0.4 − 0.6  Mo  7  0.4 + 0.4
2.2  Mo  3.2
Mo = 3

P7 (3) = C73  (0.4)  0.64  0.29

3

O.Pavlenko. Probability theory. 2020.

Example n = 8, 8  0.9 − 0.1  Mo  8  0.9 + 0.9
p = 0.9 7.1  Mo  8.1
Mo = 8

Example n = 4,
p = 0.8
4  0.8 − 0.2  Mo  4  0.8 + 0.8
3  Mo  4
Mo = 3; 4

O.Pavlenko. Probability theory. 2020.

Bernoulli formula

n−m n!
Pn (m) = C p q
m m
= p m q n−m
m!(n − m )!
n

O.Pavlenko. Probability theory. 2020.

Approximate formulas

De Moivre–Laplace local formula

de Moivre, 1730.
1  m − np 
Pn (m)     p=q=0.5
npq  npq 
P. Laplace, 1812.
x2 0<p<1
1 −
 ( x) = e 2
2
 (− x) =  ( x)

O.Pavlenko. Probability theory. 2020.

Approximate formulas
Example. In an assembly shop, on average, every eighth
electronic device is assembled inaccurately, so it needs to be
adjusted. Calculate the probability that among the 95 devices
there are exactly 10 that need to be adjusted.
1
n = 95, m = 10, p = = 0.125, q = 0.875
8
 10 − 95  0.125 
  0.31   (0.58) =
1
P95 (10) = 
95  0.125  0.875  95  0.125  0.875 
= 0.31  0.3372  0.1045

P95 (10) = C95

10
(0.125) (0.875)  0.1107
10 85

O.Pavlenko. Probability theory. 2020.

De Moivre–Laplace integral formula

 m2 − np   m1 − np 
Pn (m1  m  m2 )    −  
 npq   npq 
   

x t2
1 −
 ( x) =
2 e
−
2
dt

(− x) = 1 − ( x)

O.Pavlenko. Probability theory. 2020.

Example 250 people take part in a marathon race. Usually 8
from 10 participants on average cross the finish line.
Compute the probability that the finish line will be crossed
by
a) from 190 till 220 participants;
b) At least 215 participants.
8
n = 250, p= = 0.8, q = 0 .2
10
 220 − 250  0.8   190 − 250  0.8 
P250 (190  m  220) =   −  =
 250  0.8  0.2   250  0.8  0.2 
=  (3.16) −  (−1.58) =  (3.16) − (1 −  (1.58))
 m2 − np   m1 − np 
Pn (m1  m  m2 )    −  
 npq   npq 
   

O.Pavlenko. Probability theory. 2020.

 Example 250 people take part in a marathon race. Usually 8
from 10 participants on average cross the finish line.
Compute the probability that the finish line will be crossed
by
a) from 190 till 220 participants;
b) At least 215 participants.
8
n = 250, p= = 0.8, q = 0 .2
10
 220 − 250  0.8   190 − 250  0.8 
P250 (190  m  220) =   −   =
 250  0.8  0.2   250  0.8  0.2 
=  (3.16) −  (−1.58) =  (3.16) − (1 −  (1.58)) = 0.999 + 0.943 − 1 = 0.942

 250 − 250  0.8   215 − 250  0.8 

P250 (m  215) = P250 (215  m  250) =   −   =
 250  0.8  0.2   250  0.8  0.2 
=  (7.9) −  (2.37) = 1 − 0.991 = 0.009
O.Pavlenko. Probability theory. 2020.
Poisson formula

m
Pn (m)  e −
m!

 = np  = nq

p  0.1 p  0.9

np  10 nq  10

Pn (n − m) = ...

O.Pavlenko. Probability theory. 2020.

Example A factory sends 5000 of valid products to storehouse.
0.001 is the probability that a product will be damaged during
the time of delivery. Compute the probability that the
storehouse will receive
a) exactly two damaged products;
b) at least two damaged products.
n = 5000, p = 0.001, m=2
 = 5000 0.001 = 5
a) 5 2 −5
P5000 (2) = e  0.08
2!
b) P (m  2) = 1 − P (m  2) = 1 − P (0) − P (1) =
5000 5000 5000 5000

50 −5 51 −5
= 1 − e − e  0.96
0! 1!
O.Pavlenko. Probability theory. 2020.
 Conditions for using Bernoulli formula and
approximate formulas
1. If n is not large (n≤40) , use Bernoulli formula,

2. If n is large (n>40), p<0.1 and np<10 use Poisson formula,

or p>0.9 and nq<10 use Poisson formula,

3. If n is large (n>40), 0.1<p<0.9 or np≥10 (nq<10 ) or both last

conditions are fulfilled, use de Moivre-
Laplace formula.
If the number of successes m belongs to
an interval then use de Moivre-Laplace
integral formula,
if m is one number then use de Moivre-
Laplace local formula.

* In this course we assume that the number of experiments is large if it is greater than 40.

O.Pavlenko. Probability theory. 2020.

 Calculation of probability Pn(m) in MS EXCEL
Use EXCEL functions: BINOMDIST, POISSON, NORMDIST

Number_s is the number of

BINOMDIST successes in trials;
Trials is the number of
independent trials;
Probability_s is the
probability of success on each
trial;
Cumulative is a logical value:
for 𝑃𝑛 (𝑚 ≤ 𝑚0 ) use 1
(TRUE);
for 𝑃𝑛 (𝑚), use 0 (FALSE)

O.Pavlenko. Probability theory. 2020.

 Calculation of probability Pn(m) in MS EXCEL
Use EXCEL functions: BINOMDIST, POISSON, NORMDIST

X is the number of successes

POISSON in trials;
Mean is the mean value
λ=np;
Cumulative is a logical value:
for 𝑃𝑛 (𝑚 ≤ 𝑚0 ) use 1
(TRUE);
for 𝑃𝑛 (𝑚), use 0 (FALSE)

O.Pavlenko. Probability theory. 2020.

 Calculation of probability Pn(m) in MS EXCEL
Use EXCEL functions: BINOMDIST, POISSON, NORMDIST

X is the number of successes

NORMDIST in trials;
Mean is the mean value
λ=np;
Standard_dev is the standard
deviation  = npq
Cumulative is a logical value:
use 1 (TRUE) to compute the
value Φ(m)
n = 48,
p = 0.75,
q = 0.25

O.Pavlenko. Probability theory. 2020.