Introduction To Two-Way ANOVA

Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Introduction to Two-Way ANOVA

In a two-way analysis of variance we analyze the dependence
of a continuous response on two, cross-classied factors.
Two-Way Analysis of Variance: ANOVA

Dr. J. Kyle Roberts
Southern Methodist University Simmons School of Education and Human Development Department of Teaching and Learning
The factors can be experimental factors that are both of
interest or they can be one experimental factor and one blocking factor.
A blocking factor is a known source of variability, such as the
subject or, more generally, the experimental unit. We expect this factor to account for a substantial portion of the variability in the response. We wish to control for this variability but are not interested in comparing the particular levels of this factor.
A cross-classied experiment is said to be balanced if every
pair of conditions occurs the same number of times. You can check with the xtabs function.
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Hypothesis Test in Two-Way ANOVA

Recall that the null hypothesis for a one-way ANOVA can be
How the Data are Analyzed in a Two-Way
written as: H0 : Y 1 = Y 2 = Y 3
For the two-way ANOVA, we have the potential to test three For heuristic purposes, suppose that we have a dataset with
two independent variables each with two levels.

In a full factorial setting, we would view this data as:
separate hypothesis tests: one for each way and one for the interaction eect. We do not have to test all 3, but if we do, we refer to this as a full factorial ANOVA.
These hypotheses are:
Levels of First IV
H0:M ainA : 1. = 1. = = j. H0:M ainB : .1 = .2 = = .k H0:ABInteraction : 11 = 12 = 21 = 22 = = jk

where j are the rows and k are the columns.
a1 a2 Mean
Levels of Second IV b1 b2 11 12 X X 21 22 X X .1 .2 X X
Mean 1. X 2. X X
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
The Two-Way ANOVA Summary Table

Source Main A Main B ABInter. Error Total SS df kA 1 kB 1 dfA dfB dft dfA dfB dfAB n1 MS
SSA dfA SSB dfB SSAB dfAB SSe dfe
Two-Way ANOVA Practice

p 2
SSA SSt SSB SSt SSAB SSt
F
M SA M Se M SB M Se M SAB M Se
Fill in the Missing Values Below. Use pf to compute p.
The statistical signicance of F can be obtained by
Source Main A Main B ABinter Error Total
SS
df 5 4
MS
0.22 600 752 2 0.11 0.08 0.04
computing the F -critical value. Determining statistical signicance follows the same pattern for the t test only we have two sources of df : between and within.
For the two-way ANOVA, the df needed for testing that eect
Fill in the Missing Values Below. Use pf to compute p.
are the df due to that eect and the dferror . Therefore, the df numerator could be dierent for each eect.
Source Main A Main B ABinter Error Total
SS 80
df
MS 40
22 437
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Partitioning the Sum of Squares

J
Two-Way ANOVA Data

Read in a table using read.table which resides on this website
SSA = nK
j =1 K
)2 j. X (X .k X )2 (X j. X .k X )2 jk X (X
at http://faculty.smu.edu/kyler/courses/7311/twoway1.txt
Make sure that you include header=T > head(twoway)
1 2 3 4 5 6 gender program gre 2 1 24 2 1 27 2 1 33 2 1 25 2 1 26 2 1 30
SSB = nJ
k=1 K J
SSAB = n
k=1 j =1 K J
SSError =
k=1 j =1 i=1 K J n
jk )2 (Xijk X )2 (Xijk X
k=1 j =1 i=1
> str(twoway)
data.frame: 48 $ gender : int $ program: int $ gre : int obs. of 3 variables: 2 2 2 2 2 2 2 2 1 1 ... 1 1 1 1 1 1 1 1 1 1 ... 24 27 33 25 26 30 22 29 33 26 ...
SST otal =
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Structuring and viewing data
Program vs. Gender

> barplot(tapply(twoway$gre, list(twoway$program, twoway$gender), + mean), beside = T, col = rainbow(3))
data.frame: 48 obs. of 3 variables: $ gender : Factor w/ 2 levels "1","2": 2 2 2 2 2 2 2 2 1 1 ... $ program: Factor w/ 3 levels "1","2","3": 1 1 1 1 1 1 1 1 1 1 ... $ gre : int 24 27 33 25 26 30 22 29 33 26 ...
> table(twoway$gender, twoway$program)

1 2 3 1 8 8 8 2 8 8 8
10 0 5
15
20
25
30
> twoway$gender <- factor(twoway$gender) > twoway$program <- factor(twoway$program) > str(twoway)
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Interaction Plot
> interaction.plot(twoway$gender, twoway$program, twoway$gre)
Running the Data

> m1 <- aov(gre ~ gender + program, twoway) > summary(m1)
32
twoway$program 3 1 2
mean of twoway$gre
gender program Residuals
Df Sum Sq Mean Sq F value Pr(>F) 1 38.52 38.521 2.2942 0.1370 2 60.67 30.333 1.8066 0.1762 44 738.79 16.791
30
31
> m2 <- aov(gre ~ gender * program, twoway) > summary(m2)

Df Sum Sq Mean Sq F value Pr(>F) gender 1 38.52 38.521 2.5839 0.11544 program 2 60.67 30.333 2.0347 0.14340 gender:program 2 112.67 56.333 3.7788 0.03097 Residuals 42 626.12 14.908
27
28
29
1 twoway$gender
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Testing Assumptions
> bartlett.test(gre ~ gender * program, twoway)
Bartlett test of homogeneity of variances data: gre by gender by program Bartletts K-squared = 3.865, df = 1, p-value = 0.0493 Tables of means Grand mean 27.85417 gender gender 1 2 26.958 28.750
Investigation of Means
> model.tables(m2, "means")
> bartlett.test(gre ~ program * gender, twoway)

Bartlett test of homogeneity of variances data: gre by program by gender Bartletts K-squared = 6.5532, df = 2, p-value = 0.03776
> fligner.test(gre ~ gender * program, twoway)

Fligner-Killeen test of homogeneity of variances data: gre by gender by program Fligner-Killeen:med chi-squared = 1.6152, df = 1, p-value = 0.2038
program program 1 2 3 27.187 26.938 29.437 gender:program program gender 1 2 3 1 27.38 27.12 26.37 2 27.00 26.75 32.50
Introduction
R Exercises
Homework
Introduction
R Exercises
Homework
Post Hoc Tests

> TukeyHSD(m2)$"gender:program"
2:1-1:1 1:2-1:1 2:2-1:1 1:3-1:1 2:3-1:1 1:2-2:1 2:2-2:1 1:3-2:1 2:3-2:1 2:2-1:2 1:3-1:2 2:3-1:2 1:3-2:2 2:3-2:2 2:3-1:3 diff -0.375 -0.250 -0.625 -1.000 5.125 0.125 -0.250 -0.625 5.500 -0.375 -0.750 5.375 -0.375 5.750 6.125 lwr -6.13810348 -6.01310348 -6.38810348 -6.76310348 -0.63810348 -5.63810348 -6.01310348 -6.38810348 -0.26310348 -6.13810348 -6.51310348 -0.38810348 -6.13810348 -0.01310348 0.36189652 upr 5.388103 5.513103 5.138103 4.763103 10.888103 5.888103 5.513103 5.138103 11.263103 5.388103 5.013103 11.138103 5.388103 11.513103 11.888103 p adj 0.99995875 0.99999450 0.99949105 0.99516934 0.10654536 0.99999983 0.99999450 0.99949105 0.06898988 0.99995875 0.99876887 0.08000676 0.99995875 0.05082404 0.03144884
In-Class Assignment
Run a two-way ANOVA with the following dataset
GT
Regular
Special Ed.
Teaching Method 1 2 3 40 32 30 36 29 35 38 36 33 24 20 18 21 29 26 20 25 26 22 20 18 18 18 14 16 19 19
Introduction
R Exercises
Homework
Take-Home Assignment
1. Create a dataset of no less than 20 people with two categorical/grouping/independent variables and one dependent variable. Make the dataset a 4x2 design. Create the dataset in such a way that the main eect for the rst way is not statistically signicant, the main eect for the second way is not statistically signicant, but the interaction eect is statistically signicant. 2. Create a dataset of no less than 20 people with two categorical/grouping/independent variables and one dependent variable. Make the dataset a 4x2 design. Create the dataset in such a way that the main eect for the rst way is statistically signicant, the main eect for the second way is statistically signicant, but the interaction eect is not statistically signicant.

Introduction To Two-Way ANOVA

Uploaded by

Copyright:

Available Formats

Introduction To Two-Way ANOVA

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Introduction To Two-Way ANOVA

Uploaded by

Copyright:

Available Formats

Introduction

Introduction to Two-Way ANOVA

of a continuous response on two, cross-classied factors.

Two-Way Analysis of Variance: ANOVA

The factors can be experimental factors that are both of

Hypothesis Test in Two-Way ANOVA

How the Data are Analyzed in a Two-Way

two independent variables each with two levels.

H0:M ainA : 1. = 1. = = j. H0:M ainB : .1 = .2 = = .k H0:ABInteraction : 11 = 12 = 21 = 22 = = jk

The Two-Way ANOVA Summary Table

Two-Way ANOVA Practice

Fill in the Missing Values Below. Use pf to compute p.

The statistical signicance of F can be obtained by

Source Main A Main B ABinter Error Total

0.22 600 752 2 0.11 0.08 0.04

Fill in the Missing Values Below. Use pf to compute p.

Source Main A Main B ABinter Error Total

Partitioning the Sum of Squares

Two-Way ANOVA Data

Structuring and viewing data

Program vs. Gender

> table(twoway$gender, twoway$program)

Running the Data

gender program Residuals

> m2 <- aov(gre ~ gender * program, twoway) > summary(m2)

> bartlett.test(gre ~ program * gender, twoway)

> fligner.test(gre ~ gender * program, twoway)

Post Hoc Tests

You might also like