Slide

7-1

Probability Distributions
Random Variables: Working with
Uncertain Numbers
Slide
7-2 Random Variable
• A specification or description of a numerical result
from a random experiment
• The number is the observation of the random variable
• The random variable is the meaning of the number
• The observed value is 17 for the random variable
“Last week’s warranty returns (number of customers)”
• Examples
• Today’s stock market close
• The number of defective parts produced today
• Next quarter’s sales
• Summaries:  = mean (expected value)
 = standard deviation
Slide
7-3 Examples of Random Variables

Random Standard
Variable Mean Deviation

X = $1.40 $1.40 $0
$1 prob 0.5
Y= { $2 prob 0.5
$1.50 $0.50

$0 prob 0.98
Z= { $95 prob 0.02
$1.90 $13.30
Slide
7-4 Discrete Random Variable
• Can list all possible outcomes
• Probability Distribution
• The list of values and probabilities. Use it to compute!
• Mean (expected value) of random variable X
 = Sum of Value  Prob of Value = E(X) = XP(X)
Gives a typical or central value of the random variable
• Standard deviation of random variable X
 = Sum of (Value – )2  Probability of Value

=  ( X - ) 2
P( X )
Tells about how far from expected this random variable will be
Slide
7-5 Example: Investment Payoffs
• Probability distribution of investment payoffs
Payoff (Value) Probability
$0 0.98
$95 0.02
 = 0  0.98 + 95  0.02 = $1.90
• The expected payoff is $1.90
• A compromise between $0 (most of the time) and $95 (rarely)

 = (0–1.90)2 0.98 + (95–1.90)2 0.02 = $13.30

• Actual payoffs approximately $13.30 above or below expected
• A compromise between being $1.90 below average (most of
the time) and $93.10 above average (rarely)
Slide
7-6
Fig 7.1.1
Example: Profit Scenarios
Scenario Profit ($millions) Probability
Great $10 0.20
Good 5 0.40
OK 1 0.25
Lousy –4 0.15
0.5 Good
0.4
Probability

0.3 OK
Great
0.2 Lousy
0.1
0.0
-5 0 5 10
Profit ($millions)
Standard deviation:
$4.40 million Expected profit: $3.65 million
Slide
7-7 Binomial Distribution
• A special type of discrete random variable
• e.g., Interview 50 random customers
• How many like the new product? 0, 1, 2, …, 49, or 50
• e.g., What percent of stocks went up yesterday?
• X is binomial if it is the number of occurrences of
some event, out of n trials, provided that
• The probability  is the same for all trials, and
• The trials are independent of one another
• so that each trial brings new, independent information

• Binomial proportion or percent: p = X/n

• This is the relative frequency of the event
• e.g., if X = 35 of n = 50 people like product, p = 35/50 = 0.70
Slide
7-8 Binomial Probabilities
• A discrete distribution 0.4
=0.05, n=25
0.2
Skewed if  is close to 0
0
0 5 10 15 20 25
0.3
0.2
=0.5, n=25
Symmetric if  is 0.5 (or close) 0.1
• Approximately normal 0
0 5 10 15 20 25

0.3 =0.9, n=25

Skewed if  is close to 1 0.2
0.1
0
0 5 10 15 20 25
Slide
7-9 Binomial Mean and Std. Deviation
• Shortcut to find mean and standard deviation
quickly for binomial random variables (X or p)
• No need to compute the probability distribution
• work directly from n and 

Number of Proportion or
Occurrences, X Percent, p = X/n

Mean X = n p = 

Std. Dev. X = n(1–) p = (1–)/n

Slide
7-10 Example: Sampling Defective Parts
Draw a random sample of n=200 from the day’s production
• Suppose that  = 7% of this production is defective
Number of Proportion or Percent
Defects, X Defective, p = X/n
Mean X = n p =  = 0.07
= 2000.07 = 14
“Expect to see 14 “Expect to see 7%
defects, on average” defective, on average”

Std. Dev. X = n(1–) p = (1–)/n

= 2000.070.93 = 0.070.93/200
= 3.61 = 0.018
“Typically expect approx. “Typically expect approx.
3.61 more or less than 14” 1.8% more or less than 7%”
Slide
7-11 Computing Binomial Probabilities
• Binomial probability that X equals a
n a
P( X  a )     (1 - ) n -a
a
n!
  a (1 - ) n -a
a! ( n - a )!
1  2  3  ...  n
  a (1 - ) n -a
[1  2  3  ...  a ][1  2  3  ...  ( n - a )]
• Example: probability that exactly a = 2 of your n = 6
major customers will call tomorrow (assuming that
=0.25 is the probability that each one will call)
6
P( X  2)   0.252 (1 - 0.25) 6- 2
2
 15  0.0625  0.316406  0.297
Slide
7-12 Normal Distribution
• A special continuous distribution (not discrete)
• For every mean  and (positive) standard
deviation  there is a normal distribution
• The mean  moves the curve left and right
• The standard deviation  widens and narrows the curve

 


Slide
7-13 Probability: Area Under the Curve
• Probability of observing a value between a and b
is area under the curve
More likely
Less likely
a b a b
• Note: total area = 1
Probability = 0.50

Probability = 0.68 Probability = 0.95

(one std. dev.) (two std. devs.)

   
 
Slide
7-14 Standard Normal Distribution
• Normal with mean  = 0 and std. deviation  = 1

-3 -2 -1 0 1 2 3
• Standard normal probability table
• Gives probability that a standard normal is less than a
given value
Probability Examples
= 0.3085 Value Probability
–1 0.1587
0 0.5
-3 -2 -1 0 1 2 3 1 0.8413
Value = –0.5 2 0.9772
Slide
7-15 Finding Normal Probabilities
• Example: Sales are forecast as $80 million (mean) with a standard
deviation of $10 million. Find the probability that sales will exceed
$86 million, assuming a normal distribution
• Figure out the question
• Find Prob(X>86) where =80 and  = 10
• Standardize (Subtract , divide by  to get std. normal)
X– 86–80
• Prob ( >
10
)
Standard
= Prob (
normal > 0.60)
• Draw Picture
-3 -2 -1 0 1 2 3
0.60

• Use tables, find answer: 1–0.7257 = 0.274

Slide
7-16 Normal Approx. to the Binomial
• Idea: to make it easier to compute binomial
probabilities
• If
• n is large, and
•  is not too close to 0 or 1,
• Then
• Binomial probabilities for X are close to normal
probabilities with
•  = X = n •  = X = n(1–)
• Similarly for p=X/n with
•  = p =  •  = p = (1–)/n
Slide
7-17 Example: an Opinion Poll
• Suppose  = 55% approve president’s performance
• We interview n = 400 chosen at random
• What are the chances that fewer than 200 will say they
approve? That is, find Prob (X < 200)
• When you round any number less than 199.5 to the nearest
whole number, the result will be “fewer than 200”
• The mean is X = n = 4000.55 = 220
• The standard deviation is X = n(1–) = 9.9499
• We need to compute the probability that a normal random
variable with this mean and this standard deviation is less than
199.5
X–X 199.5–220 Standard
Prob ( 
<
9.9499
) = Prob ( )
normal < –2.06
= 0.020
Slide
7-18 Example: Poll (continued)
• Still assuming 55% approve and we interview 400
at random
• What are the chances that more than 58% will say they
approve?
• Convert from percentage p to number of people X
• Note that 58% of 400 people is 0.58400 = 232 people
• Numbers that round to “more than 232” are 232.5 and above
• We need to compute the probability that a normal random
variable with mean X = 220 and standard deviation X =
9.9499 is more than 232.5
X–X 232.5–220 Standard
Prob ( 
>
9.9499
) = Prob ( normal > 1.26 ) = 0.10
• About 10% of the time we will find more than 58%