You are on page 1of 26

ANOVA - ANalysis

Of VAriance
INTRODUCTION
Jacob’s chemical company obtained data to
test whether the mean time to mix a batch of
material is the same for machines produced
by 3 manufacturers. The data obtained was
in time (in min) needed to mix the material.
Use these data to test whether the
population mean times for mixing batch of
material differ for three manufacturers. Use
alpha = .05
INTRODUCTION
Manufacturer
First Second Third
20 28 20
26 26 19
24 31 23
22 27 22
Average 23 28 21
Variance 6.67 4.67 3.33
Std Dev 2.58 2.16 1.83
INTRODUCTION
Let µ1 = mean population time for machine 1

• µ2 = mean population time for machine 2

• µ3 = mean population time for machine 3

• Ho : µ1 = µ2 = µ3

• H1 : Not all population means are equal


INTRODUCTION
Anova can be analyzed from data obtained
by two types of study: Observational study
and Experimental study

Common terminology: Response variables,


Factors and Treatments

Response Variables: These are dependent


variables like here the time taken by each
m/c is a response variable
INTRODUCTION

Factors: Independent variables like in this


example, machines are factors

Treatments: The three levels of factors. First,


Second and Third m/c are 3 treatments
ASSUMPTIONS FOR
ANOVA
For each population, the response variable is
normally distributed: In this example, time
taken by each m/c is normally distributed

The variance of the response variable σ2 is


same for all the population

The observations must be independent: In


this example, the time taken by first m/c is
independent of second m/c
ASSUMPTIONS FOR
ANOVA
If null hypothesis was true, i.e. µ1 = µ2 = µ3,
we can use the variability of sample means to
develop an estimate of σ2 .

If the assumption on variance was satisfied,


then the sample means must have come
from same normal distribution with a mean µ
and std. dev σ
CONCEPTUAL
OVERVIEW

σxbar = σ2 / n

x1 µ x2 x3
Conceptual Overview

If null hypothesis was true, then x1 bar = 23,


x2 bar = 28 and x3 bar = 21 must have drawn
from same normal distribution

We can use sample mean and std dev to


estimate the mean and variance of the
sampling distribution
Conceptual Overview
Estimate of the mean of sampling distribution = (23
+ 28 + 21) / 3 = 24

Estimate of the variance of sampling distribution is


given by
• sx2 = (23 - 24)2 + (28 - 24)2 + (21 - 24)2 / (3 - 1)
• = 13

• Since σxbar = σ2 / n; Estimator ofσ2 = n (Estimator of σxbar )


= 4 * 13 = 52 is called the between-treatments estimate of
σ2
Conceptual Overview
When a simple random sample is selected from a
population, each of the sample variance provides
an estimator of σ2. So, these variances can be pooled
into one and that is called as within-treatments estimate of
σ2. This is obtained by average of individual sample
variances = (6.67 + 4.67 + 3.33) / 3 = 4.89

Now the ratio of between-treatments variance and within-


treatments variance = 52 / 4.89 = 10.64
Testing for equality of ‘k’
population means

H0 = µ1 = µ2 = ........= µn

• H1 = Not all population means are equal


Between-treatments
• This estimate variance
of σ2 is called as ‘Mean Square
Due to Treatments’ i.e. MSTR

• MSTR = SSTR (Sum of Squares Due to


Treatments) / (k-1)

• SSTR = ∑ nj (xjbar - xdoublebar)

• SSTR = 4(23 - 24) 2 + 4(28 - 24)2 + 4(21 - 2


24) = 4
+ 64 + 36 = 104

• MSTR = SSTR / (k - 1) = 104 / 2 = 52


Within-treatments variance
• This estimate of σ2 is called as ‘Mean Square
Due to Error’ i.e. MSE

• MSE = SSE (Sum of Squares Due to Error) / (nT -


k)

• SSE = ∑ (nj - 1) sj2

• SSE = (4 - 1) 6.67 + (4 - 1) 4.67 + (4 - 1) 3.33 =


44

• MSE = SSE / (nT - k) = 44 / (12 - 3) = 44 / 9 = 4.89


ANOVA TABLE

Source Degrees
Sum of Mean
of of F p-value
Squares Square
variation Freedom
Treatme
104 2 52 10.64
nts

Error 44 9 4.89

Total 148 11
• EXERCISE
A study investigated the perception of corporate
ethical values among individuals specializing in
marketing. Use alpha =.05, and the below given
data (higher scores indicate higher ethical values) to
test for significant differences in mean perception
among the three groups.
Marketing mgrs Marketing Research Advertising
6 5 6
5 5 7
4 4 6
5 4 5
6 5 6
4 4 6
TWO-WAY ANOVA
TWO-WAY ANOVA

• Main Effect (Factor A): Preparation


program
• Main Effect (Factor B): Undergraduate
program
• Interaction Effect (Factors A & B)
• SST = SSA + SSB + SSAB + SSE
TWO-WAY ANOVA
TWO-WAY ANOVA
TWO-WAY ANOVA
FISHER’S LEAST SIGNIFICANT
DIFFERENCE

• FISHER’S LSD PROCEDURE

• H0: µ1 = µ2

• H1: µ1 ≠ µ2

• Test statistic: t = (xi bar - xj bar) / MSE (1/n1 + 1/n2)

• Reject if p-value ≤ alpha


FISHER’S LEAST SIGNIFICANT
DIFFERENCE
• FISHER’S LSD PROCEDURE

• H0: µ1 = µ2

• H1: µ1 ≠ µ2

• Test statistic: t = (xi bar - xj bar)

• Least Significant Difference = tα/2 MSE (1/ni + 1/nj)

• Reject if | xi bar - xj bar | > Least Significant


Difference
Randomized Block Design - 2 Way
ANOVA
• An automobile dealer conducted a test to determine if the time
in minutes needed to complete a minor engine tune-up depends
on whether a computerized engine analyzer or an electronic
analyzer is used. Because tune-up time varies among compact,
intermediate, and full-sized cars, the 3 types of cars were used
as blocks in the experiment. At α=0.05, test for any significant
difference. The data obtained follow:
Analyzer
Computerized Electronic
Compact 50 42
Car Intermediate 55 44
Full-sized 63 46
RANDOMIZED BLOCK DESIGN - 2
WAY ANOVA
Degrees
Source of Sum of
of Mean Square F
variation Squares
Freedom
MSTR = MSTR/MS
Treatments SSTR k-1
SSTR / k - 1 E
MSBL =
Blocks SSBL b-1
SSBL / b - 1
MSE =
Error SSE (k-1) (b-1) SSE / (k-1)
(b-1)

Total SST nT - 1

You might also like