Today’s topics: Chapters 11

(1) Continuous Random variables

(2) Normal Distribution: Standard normal distribution;

Lec 11—Chapter 11 Standardization

Yijun Lou
(3) Approximation to Binomial distribution by Normal
November 21, 2016 Distribution

(4) Sampling distributions and estimation

2/33

Review of two discrete distributions 11.2. Continuous Random Variable

X ≥ B (n, p): n–number of trials, p–the probability
2.1 Probability Distribution
of success in one trial: The probability of k successes
A B Suppose the random variable takes values on the in-
n k
P (k) = P (X = k) = p (1 ≠ p)n≠k . terval (c,d) or [c,d), (c,d], [c,d]. Then the probability
k
distribution is assigned by a function f (x), x œ (c, d)
(or other intervals). The function f (x) is called the
µ = E(X) = np and ‡ 2 = V ar(X) = np(1 ≠ p). “probability density function" (pdf):

(a) The total area under this curve bounded by the x

Poisson Distribution: The number of occurrences in axis is equal to one.
an interval X ≥ Pois(⁄). The probability of k occur- (b) The area under the curve between lines x = a and
rences x = b gives the probability that X lies between
⁄k e≠⁄
P (k) = P (X = k) = a and b. That is P (a Æ X Æ b)=area.
k!
where ⁄ is the mean number of occurrences. (c) f (x) Ø 0 for all x.
µ = E(X) = ⁄ and ‡ 2 = V ar(X) = ⁄.

3/33 4/33
 Definition 11.2.2: Let X be a continuous random
An important notation: Integration of f (x) from a
variable with probability density distribution f (x).
to b (a < b)-The area under the curve between lines
x = a and x = b (1) The mean or expected value:
⁄ b ⁄ Œ
f (x)dx. µ = E (X) = xf (x) dx
a ≠Œ

Remarks on probabilities: (2) The variance of X:

sa ⁄ Œ
(a) P (X = a) = a f (x)dx = 0, the probability that 2
‡ = V ar (X) = (x ≠ µ)2 f (x) dx
a continuous randmon variable will assume a ≠Œ
fixed value is zero
Alternative formula
s
(b) P (a Æ X Æ b) = ab f (x)dx is the probabil- 2
⁄ Œ
‡ = V ar (X) = x2 f (x) dx ≠ µ2
ity that the continuous random variable X lies ≠Œ
between a and b.
sa (3) The standard deviation of X:
(c) P (X < a) = P (x Æ a) = ≠Œ f (x)dx: probabil- Ò
ity that X Æ a. ‡= V ar(X)
Remark: This page is just for your interest.

5/33 6/33

11.2.2 Normal Distribution Remark on the normal distribution:

Definition 11.2.3: A continuous random variable X is

(1) Notation X ≥ N (µ, ‡ 2 ): X follows a normal
defined to be a normal random variable if its probability
distribution with mean µ and variance ‡ 2 .
function is given by
1 2 2
(2) It is a continuous distribution.
f (x) = Ô e≠(x≠µ) /2‡
2fi‡ (3) The curve is symmetric and bell-shaped about a
where µ is the mean of X and ‡ is the standard vertical axis through the mean µ.
deviation of X.
(4) The total area under the curve and above the hori-
sŒ
Remark: This page is just for your interest. You do not zontal axis is equal to 1. That is ≠Œ f (x)dx =
need to memorize the pdf for normal distribution. 1.

7/33 8/33
11.2.3 Standard Normal Distribution Example: Given that Z ≥ N (0, 1), find the following
probabilities by using the standard normal distribution
table
Definition 11.2.6: The normal distribution with µ = 0
and ‡ = 1 is called the standard normal distribution. (a) P (Z > 2.48)
Denoted as Z ≥ N (0, 1) if Z is a standard normal
random variable. (b) P (Z < 0.92)

(c) P (0 Æ Z Æ 1.84)
Standard Normal Distribution table for P (Z > z) (d) P (≠0.58 Æ Z Æ 1.36)

Solution: (a) P (Z > 2.48)=0.00657.

9/33 10/33

(b) P (Z < 0.92) = 1 ≠ P (Z Ø 0.92) (c) P (0 Æ Z Æ 1.84) = P (Z > 0) ≠ P (Z > 1.84)

= 1 ≠ 0.1788 = 0.5 ≠ 0.0329
= 0.8212. = 0.4671.

11/33 12/33
 Standardizing a normal distribution

(d) P (≠0.58 Æ Z Æ 1.36) Let X ≥ N (µ, ‡ 2 ) (X is a normal random variable

= 1 ≠ P (Z < ≠0.58) ≠ P (Z > 1.36) with mean µ and standard deviation ‡), then the
= 1 ≠ P (Z > 0.58) ≠ P (Z > 1.36) variable
= 1 ≠ 0.2810 ≠ 0.0869 Z = (X ≠ µ)/‡

= 0.6321. is a standard normal variable, that is

Z ≥ N (0, 1).

Hence,
x1 ≠ µ x2 ≠ µ
P (x1 < X < x2 ) = P ( <Z< )
‡ ‡

13/33 14/33

Example: Suppose X ≥ N (30, 102 ), find the probabil-

ities (a) P (30 Æ X Æ 49) and (b) P (25 Æ X Æ 38).
(a) P (30 Æ X Æ 49)
30 ≠ 30 X ≠ 30 49 ≠ 30
= P( Æ Æ )
Solution: X ≥ N (30, 102 ), then 10 10 10
= P (0 Æ Z Æ 1.9)
X ≠ 30
Z = (X ≠ µ)/‡ = = P (Z > 0) ≠ P (Z > 1.9)
10
= 0.5 ≠ 0.0287 = 0.4713.
follows N (0, 1)

15/33 16/33
 Motivation: Determining the z0 values in the Normal
Distribution given the probability is known.
(b) P (25 Æ X Æ 38)
25 ≠ 30 X ≠ 30 38 ≠ 30
= P( Æ Æ ) Example 11.2.14: Find the value of z0 such that
10 10 10
= P (≠0.5 Æ Z Æ 0.8) P (Z > z0 ) = 0.005.

= 1 ≠ P (Z < ≠0.5) ≠ P (Z > 0.8)

Solution: Directly check the standard normal distribution
= 1 ≠ P (Z > 0.5) ≠ P (Z > 0.8) table, we have
= 1 ≠ 0.3085 ≠ 0.2119
P (Z > 2.57) = 0.00508 and P (Z > 2.58) = 0.00494.
= 0.4796.
Hence, we may choose z0 as either 2.57 or 2.58.

17/33 18/33

Example 11.2.15: Find the value of z0 such that Motivation: Suppose X ≥ N (µ, ‡ 2 ). That is X fol-
P (Z < z0 ) = 0.005. lows a normal distribution with mean µ and variance
‡ 2 . Determining the x0 values in the Normal Distribu-
Solution: Since P (Z < 0) = 0.5 while 0.005 < 0.5, we tion given the probability is known (for example, sup-
have z0 < 0. However, pose X ≥ N (10, 4), find x0 in P (X > x0 ) = 0.02).

P (Z < z0 ) = P (Z > ≠z0 ) = 0.005. Main idea: Use

Based on example 11.2.14, we have ≠z0 = 2.57 or 2.58. Z=

X ≠µ
≥ N (0, 1)
Therefore, we may choose z0 as either ≠2.57 or ≠2.58. ‡

19/33 20/33
 Example: Suppose X ≥ N (200, 102 ), find x0 such
that P (X < x0 ) = 0.5%. Example 11.2.21: Assume that X follows a binomial
distribution with n=100 and p = 0.36. Find the
Solution: Let Z = 10 ,
X≠200
then Z ≥ N (0, 1) and following probability:
x0 ≠ 200 P (X > 51), P (X Ø 51)
P (X < x0 ) = P (Z < z0 ) = P (Z < ) = 0.005.
10
We choose z0 = ≠2.58 (based on Example 11.2.15). Idea: Normal distribution Y ≥ N (µ, ‡ 2 ) as an approxi-
Therefore, mation to the binomial distribution X ≥ B(n, p)
x0 ≠ 200
= ≠2.58
10
and
x0 = 200 + (10)(≠2.58) = 174.2.

21/33 22/33

Normal distribution Y ≥ N (µ, ‡ 2 ) as an Solution to Example 11.2.21:

approximation to the binomial distribution
X ≥ B(n, p) Step 1: np = 36 and n(1 ≠ p) = 64. Both are greater
than 5 and the approximation works
Step 1 Check whether the approximation work np Ø 5
and n(1 ≠ p) Ø 5 Step 2: µ=np=36 and ‡ 2 =np (1 ≠ p)=36 ú 0.64. We
use the normal approximation of binomial distribution.
Step 2 Compute µ and ‡ for the binomial distribution That is,
Y ≥ N (100(0.36), 36 ú 0.64).
µ = np and ‡ 2 = np (1 ≠ p)
Step 3:

Step 3 Correction for continuity: (a) P (X = a) ¥ a. P (X > 51) = P (Y > 51.5)

3 4
P (a ≠ 0.5 < Y < a + 0.5); (b) P (X Ø a) ¥ Y ≠µ 51.5 ≠ 36
= P > Ô
P (Y > a≠0.5); (c) P (X > a) ¥ P (Y > a+0.5); ‡ 23.04
(d) P (X Æ a) ¥ P (Y < a + 0.5); (c) P (X < = P (Z > 3.22)
a) ¥ P (Y < a ≠ 0.5) = 0.00064

23/33 24/33
 Sampling Distributions and Estimation

12.2.1 Sampling distribution of the sample

mean
b. P (X > 51) = P (Y > 50.5)
3
Y ≠µ 50.5 ≠ 36
4 Sample vs. Population
= P > Ô
‡ 23.04
= P (Z > 3.02) The State of Work-Life Balance in Hong Kong
= 0.00126
(1) Aim: Want to know the state of Work-Life Balance
for all employees (for example, more than 3 million)
in HK (Population data, such as mean and
variance)

(2) How to do the survey (sampling): Take 1000 people

for the sampling (Sample data) to get the sample
mean and sample variance

(3) Use sample mean and sample variance to infer

the values for Population mean and variance

25/33 26/33

Motivation example: Suppose HONDA service center Example: A HONDA service center charges $40, $45
wants to estimate its average revenue for randomly and $50 for tune-ups for 4, 6 and 8 cylinder engines,
selected tune ups. It currently charges $40, $45 and respectively. If 20%, 30% and 50% of its tune-ups
$50 to service 4, 6 and 8 cylinder engines. How does are done for 4, 6 and 8 cylinder cars, find (a) the
it go about doing this. probability distribution of revenue for a single randomly
selected tune-up; (b) the mean and variance of revenue.
Answer: Let X be a random variable of venue for
(Do this question as an exercise)
a randomly selected tune up of a car from HONDA.
Then Honda needs the probability distribution for X, Solution: Let X be random variable of venue for a
or it needs the population distribution. randomly selected tune up of a car from HONDA. Then
(a)
X $40 $45 $50
P (x) ? ? ?
X $40 $45 $50
P (x) 0.2 0.3 0.5
Definition 12.1.1: The population distribution is the
probability distribution of the population data.

27/33 28/33
(b) Question: If HONDA doesnot actually have the P (x)
for random variable X above, how does it estimate µ
µ = E[X] and ‡ 2 ?
ÿ
= xP (x)
x
One may do the following steps:
= 40(0.2) + 45(0.3) + 50(0.5)
= 46.5. 1. Step 1: Take samples to compute sample mean
ÿ
‡ 2
= 2
(x ≠ µ) P (x) = 15.25 and sample variance
x
2. Step 2: Infer information on population statis-
tics (say µ and ‡ 2 ) from samples (sample mean
and sample variance)

29/33 30/33

Throughout this lecture, we always use N to denote the

population size, while n denotes the sample size.
Example: Suppose on a particular day, only 2 servicing
Definition: Let X1 , X2 , . . . , Xn be a random sample jobs involve a tune-up ($40, $45 and $50 to service
from random variable X, then 4, 6 and 8 cylinder engines). Let X1 be revenue from
1ÿ n first tune-up and X2 be revenue from second tune-up.
X= Xi Find the sampling distributions of
n i=1
X1 + X2
is called the sample mean, which is also a random X=
2
variable, denoted by X. The probability distribution
of X is called the sampling distribution of X. the sample average revenue.

Definition: The mean and standard deviation calcu-

lated for the sampling distribution of X are called the
mean µX and standard deviation ‡X of X.

31/33 32/33
Solution: The following table lists all possible (X1 , X2 )
pairs resulting values of X and s2 :

X1 X2 X s2
40 40 40 0
40 45 42.5 12.5
40 50 45 50
45 40 42.5 12.5
45 45 45 0
45 50 47.5 12.5
50 40 45 50
50 45 47.5 12.5
50 50 50 0

Sampling distribution of X investigates the distribution

of these numbers for X. 1
1
Remark: The sample variance s2 is presented in Definition
12.3.1.

33/33