Sei sulla pagina 1di 8

TOM 302

LN 1B-1
Jan/15

Basic Probability Theory (Chapter 5)


Commonly associated with attributes data
Probability is a number between 0 1
0 or 1 are certainties, not of an environment of assignable risk
P = 0 is an event that can never occur: not possible
P = 1 is an event that always happens: sure thing
P (event)

# of ways an event can occur


total # of possible outcomes

Two types of probability


Subjective more likely than not type of opinions (e.g. weather)
Objective calculated by unbiased methods
Two types of objective probabilities
1. A Priori
by the nature of the event
Examples:
coin, die, container of numbered ping-pong balls (Lottery)
2. Empirical observed events
of 10,000 people age 30 years
9,800 survive to age 35 years
p (5 year survival for a 30 year old) = .98
[does not mean an individual will be 98% alive at age 35]
Five Probability Calculation Rules
(Formulas and explanatory diagrams)
1. COMPLEMENTATION LAW (not A)
P (not A) = 1 P (A)

u universe; p (u) = 1.00


Venn diagram
2. ADDITION LAW (OR for mutually exclusive events)
P (A or B) = P (A) + P (B)
events A, B are separate
3. ADDITION LAW (OR for events that are not m. e.)
P (A or B) = P (A) + P (B) P (A&B)
events A, B have overlap
4. MULTIPLICATION LAW (AND independent events)
P (A and B) = P (A) x P (B)
urn with 6 blue, 3 red
5. MULTIPLICATION LAW (AND dependent events)
P (A and B) = P (A) x P (B A)
example P (red & red)
Also
= P (B) x P (A B)
w/o replacement
Conditional probability P (B A) reads: probability of B given that A has occurred.

TOM 302

LN 1B-2

Review Populations and Samples


Measure
Size
Data element
Proportion
Mean
Standard Deviation
Variance

Population (parameters)
N
xi

Sample (statistics)
n
xi
p
x-bar
s
s2

BiNomial Probabilities and Discrete Probability Distributions (Chapters 5 & 6)


p probability of an event happening in the Popn ()
q probability that the event doesnt happen (complement, or 1 p)
n number of trials (or samples)
Basic Binomial Formula
(p + q)n

True for all n, here are some expansions


n
expansion
1
(p + q) = 1
2
p2 + 2pq + q2 = 1
3
p3 + 3p2q + 3pq2 + q3 = 1
4
etc.
There is a pattern here for the terms in the BiNomial Expansion
Each term has a coefficient (C) and the variables p, q raised to exponential powers
Cpxq(n-x)
The first term has x = n and decreases, integer-wise until x = 0
i.e. p0 = 1
(any variable raised to the zero power = 1; from the rule of
division where the power of the denominator is subtracted from the power of a like
variable in the numerator; give examples)
The coefficient is also found from the position of the term in the expansion
sequence; one way is by use of Pascals Triangle
1
1
1
1
2
1
1
3
3
1
1
4
(let students help continue a few more lines up to 5)

TOM 302

LN 1B-3

n = 4: the coefficients are

1, 4, 6, 4, 1

and the expansion equation is: (p + q)4 = p4 + 4p3q + 6p2q2 + 4pq3 + q4 = 1


Students do the expansion for n = 6 at their desk and discuss
(1, 5, 10, 10, 5, 1 for n = 5, so Cs are 1, 6, 15, 20, 15, 6, 1 for n = 6)
(p + q) 6 =
=1
Lets look at the applications of the BiNomial expansion; it represents a discrete
probability distribution
Outcomes of n = 4 coin flips
4
HHHH

1 way

# of Heads in 4 flips
3
2
HHHT
HHTT
HHTH
HTHT
HTHH
HTTH
THHH
THHT
THTH
TTHH
4 ways
6 ways

1
HTTT
THTT
TTHT
TTTH

0
TTTT

4 ways

1 way

The # or ways, or combinations are like the coefficients of the binomial expansion
The probabilities can be calculated from the above table as follows
With 16 total combinations
The probability of the outcome x = 4 heads is 1/16 = .0625
(binomial probabilities are customarily specified to 4 decimal places)
P(x = 3 heads) is 4/16 = .2500
P(x = 2 heads) is 6/16 = .3750
P(x = 1 head) is 4/16 = .2500
P(x = 0 heads) is 1/16 = .0625
Draw the histogram
We can use the Basic Probability Laws of complementation, addition and multiplication
on these outcomes (note the entire distribution = 1.000)
Relate to a sampling scheme to check for a fair coin based on 4 flips
Fair if x > 1
Not Fair coin if x = 0 heads (although there is a probability of making an error)

TOM 302

LN 1B-4

We also could calculate any one of these binomial probabilities without constructing the
entire distribution by use of the BiNomial Probability Formula
P(x) = Cn px q(n-x)
x

(edit note: the Coefficient should have n superscript and x subscript)


Cn
x

factorials

n!
x! (n-x)!

n! = n(n-1)(n-2)1

Definition: 0! 1

give examples of calculating the coefficients of n = 4, x = 0, 1, 2, 3, 4 heads


calculate each outcome probability for = .5
(infinite Population)
n=4
P(0 heads) = (setup the calculation) = (1)(.5)0(.5)4 = (1)(1)(.0625) = .0625
P(1 head) = (setup the calculation) = (4)(.5)1(.5)3 = (4)(.5)(.1250) = . 2500
P(x=2) = (setup the calculation) = (6)(.5)2(.5)2 = (6)(.25)(.25) = .3750
P(x=3) = (setup the calculation) = (4)(.5)3(.5)1 = (4)(.1250)(.5) = .2500
P(x=4) =(setup the calculation) = (1)(.5)4(.5)0 = (1)(.0625)(1) = .0625

use of Tables (Appendix A.1 & A.2-cummulative probabilities pages A-1, A-8)
Using the cumulative table
P(x=3) = P(x<3) P(x<2) = .9375 - .6875 = .2500

TOM 302

LN 1B-5

Example of applications in Discrete Data Distributions (Bi Nomial)


Say a class has N = 25 students with the number of females in this population, x = 8.
= (8/25) = .32;
or p = .32
q = .68
P(x) = Cn px q(n-x)
x

(edit note: the Coefficient should have n superscript and x subscript)


Where:
Cn
=
n!
x
x! (n-x)!
In samples of n = 4, we can expect the following probabilities
P(x=0 females in a sample of 4) = (setup the calculation) = (1)(.32)0(.68)4 = .2138
P(x=1) = (setup the calculation) = (4) (.32)1 (.68)3 = (4)(.32)(.3144) = .4025
P(x=2) = (setup the calculation) = (6) (.32)2 (.68)2 = (6)(.1024)(.4624) = .2841
P(x=3) = (setup the calculation) = (4) (.32)3 (.68)1 = (4)(.0328)(.68) = .0891
P(4) =(setup the calculation) = (1) (.32)4 (.68)0 = (1)(.0105)(1) = .0105
Draw the distribution

P(x)

.6
.5
.4
.3
.2
.1
0

= .32
n=4
.4025
.2841
.2138
.0891
.0105
0
1
2
3
4
x, number of females in the sample

0
5

Compare with Binomial Table A.1 in the text (page A-1, = .3


There is no = .32)

TOM 302

LN 1B-6

Continuous Distributions/The Normal Distribution


Discuss the Model
Continuous for all real x from - xi +
Mean = Median = Mode = 0
Standard Deviation = Variance = 1
Symmetrical
Covers all probabilities such that A = 1.0000
So special that the random variable is given a special symbol, z
(draw the bell shaped curve and discuss)
Finding probabilities using the Normal Distribution Tables
See Inside Back Cover soft pages
Pick and do some in class exercise examples
(COMMENT: Z values to two decimal place accuracy; probabilities are customarily to 4
decimal place accuracy)
P(-1.00 < z < 1.00) = 2 x .3413 = .6826
P(-2.00 < z < 2.00) = 2 x .4772 = .9544
P(-3.00 < z < 3.00) = 2 x .4987 = .9974
P(2.00 < z < 3.00) = .9987 - .9772 = .0215
P(-3.00 < z < -2.00) = (same)
P( z > 0.95) = 1 - .8289 = .1711
It ALWAYS helps to sketch a picture and show the area to be solved
Using the Table in Reverse
P(-A < z < +A) = approx .7500
Half area = .3750
(tail would be .5000 - .3750 = .1250)

A = 1.15

P (A < z) = .5500
A must be on the negative side of z = 0 since P (0 < z) = .5000
Draw sketch with a lower tail of .4500
From the table, A = -0.13 (using -0.12 gives P(A < z ) = .5478; which is
insufficient)

TOM 302

LN 1B-7

Z Transform and applications


Converting from the x domain to z
Z=

x - x
x

units will cancel and the result will be dimensionless (no units)
Examples:
1. Rope cutting process
x = 10.0 inch
x = 0.1 inch
a. What is the probability that a cut piece is 9.9 inch < x < 10.1 inch?
Z (calculated) = 1.00
P = .6826
b. If rope cut to less that 9.9 is scrapped, what proportion of cut ropes are thrown
away?
Z = -1.00
P (-< z) = .1587
2. Filling juice cans example:
= 12 oz.
= 0.2 oz
FTC says L = 11.7 oz, or that x 11.7 oz
Draw the situation (normal distribution; show mean, sigma and L)
Find the proportion of underfilled containers
ZL = (11.7 oz 12.0 oz)/ 0.2 oz
= -1.50
from Table, P = .07 or close to 10%
Try fixing
decrease may be too expensive
= L + 3 = 11.7 oz = 3(0.2 oz)
= 12.3 oz, new process center to make P = 0%

TOM 302

LN 1B-8

3. Ability to meet specifications


Specification U, L = 2.050 + 0.02 inch
Rework > U Scrap < L
From process study
Estimate of = 2.067 in.
Estimate of = .0078 in.
Draw the distribution and limits; calculate the z scores
ZL = (2.03 2.067)/.0078 = -4.74
scrap = 0 %
ZU = (2.07-2.067)/.0078 = .38
rework = 1.0000 - .6480 = .3520 = 35%
Discuss and analyze
Assign Dice Toss Sheet (Homework Exercise available on Blackboard)
Objective is to determine what is the outcome (numerical value) of tossing two dice?
Analysis is to assemble a distribution of data by summing two dice after rolling
Data is collected by observing samples and recording results.
Due

Potrebbero piacerti anche