Introduction to 2-way ANOVA

Statistics
Spring 2005
Terminology
 2-Way ANOVA means
 2 independent variables
 1 dependent variable
 3X4 ANOVA means
 2 independent variables
 1 dependent variable
 one IV has 3 levels
 one IV has 4 levels
 HYPOTHESES TESTED
in 2-WAY ANOVA
 No differences for IV #1 (A - 3 levels) These are
 H0: MA1 = MA2 = MA3 called
“Main
 No differences for IV #2 (B - 4 levels) Effects”
 H0: MB1 = MB2 = MB3 = MB4
 No interaction
 At least one MAiBj  MAmBn
EXAMPLE
 One might suspect that level of
education and gender both have
significant impacts on salary. Using the
data found in
Census90 condensed.sav
determine if this statement is true.

Dependent Variable Independent Variables

INCOME GENDER (2 levels)
(ratio level data) EDUCAT (6 levels)
 = .05
 HYPOTHESES TESTED
for a 2X6 ANOVA

No differences for GENDER (2 levels)

H0: MMale = MFemale
No differences for EDUCATION (6 levels)
H0: MB1 = MB2 = MB3 = MB4 = MB5 = MB6
No interaction
At least one MAiBj  MAmBn
To run the test of these hypotheses in SPSS…..
Analyze  General Linear Model  Univariate

NOTE: Use
this method of
analysis when
both IV’s are
not repeated
measures.
 Estimated Marginal Means of Wage or salary incom
70000

60000

50000
Estimated Marginal Means

40000

30000

20000 Sex

10000 Male

0 Female
<= 9th grade HS diploma Bachelors Degree
HS - no diploma Some college Graduate Degree

EDUCAT
 Tests of Between-Subjects Effects

Dependent Variable: Wage or salary income, 1989

Type III Sum
Source of Squares df Mean Square F Sig.
Corrected Model 5.946E+10a 11 5405064664 12.887 .000
Intercept 1.781E+11 1 1.781E+11 424.577 .000
SEX 1.214E+10 1 1.214E+10 28.951 .000
EDUCAT 2.884E+10 5 5768187295 13.752 .000
SEX * EDUCAT 6215428787 5 1243085757 2.964 .012
Error 1.976E+11 471 419428810.4
Total 4.921E+11 483
Corrected Total 2.570E+11 482
a. R Squared = .231 (Adjusted R Squared = .213)
 Tests of Between-Subjects Effects

Dependent Variable: Wage or salary income, 1989

Type III Sum
Source of Squares df Mean Square F Sig.
Corrected Model 5.946E+10a 11 5405064664 12.887 .000
Intercept 1.781E+11 1 1.781E+11 424.577 .000
SEX 1.214E+10 1 1.214E+10 28.951 .000
EDUCAT 2.884E+10 5 5768187295 13.752 .000
SEX * EDUCAT 6215428787 5 1243085757 2.964 .012
Error 1.976E+11 471 419428810.4
Total 4.921E+11 483
Corrected Total 2.570E+11 482
a. R Squared = .231 (Adjusted R Squared = .213)

Table 1
Results of the 2-way ANOVA (Gender X Education) for Income.
Source SS df MS F p
Gender 12,142,860,574 1 12,142,860,574 28.95 0.000
Education 28,840,936,475 5 5,768,187,295 13.75 0.000
Gender X Education 6,215,428,787 5 1,243,085,757 2.96 0.012
Residual 197,550,969,717 471 419,428,810
 HYPOTHESES TESTED
for a 2X6 ANOVA
No differences for GENDER
(2 levels)
H0: MMale = MFemale
Reject H0
No differences for EDUCATION
(F(1,471)=29.95: p=.000) (6 levels)
H0: MB1 = MB2 = MB3 = MB4 =
Reject H0 MB5 = MB6
(F(5,471)=13.75: p=.000) No interaction
At least one MAiBj  MAmBn

Reject H0
(F(5,471)=2.96: p=.012)
Types of 2-Way ANOVA designs

 Both IV’s are between subjects

(i.e. not-repeated measures)
 Both IV’s are within subjects
(i.e. repeated measures)
 One IV is between subjects, the other IV is
within subjects
 Both IV’s are between subjects
(i.e. not-repeated measures)
Analyze  General Linear Model  Univariate
 Both IV’s are within subjects
(i.e. repeated measures)
Analyze  General Linear Model  Repeated Measures
Analyze  General Linear Model
 Repeated Measures
 Estimated Marginal Means of MEASURE_1
41.8

41.6

41.4
Estimated Marginal Means

41.2

DAY
41.0
1

40.8 2
1 2

TRIAL
 Tests of Within-Subjects Effects

Measure: MEASURE_1
Type III Sum
Source of Squares df Mean Square F Sig.
DAY Sphericity Assumed .443 1 .443 .214 .647
Greenhouse-Geisser .443 1.000 .443 .214 .647
Huynh-Feldt .443 1.000 .443 .214 .647
Lower-bound .443 1.000 .443 .214 .647
Error(DAY) Sphericity Assumed 64.123 31 2.068
Greenhouse-Geisser 64.123 31.000 2.068
Huynh-Feldt 64.123 31.000 2.068
Lower-bound 64.123 31.000 2.068
TRIAL Sphericity Assumed 2.872 1 2.872 1.971 .170
Greenhouse-Geisser 2.872 1.000 2.872 1.971 .170
Huynh-Feldt 2.872 1.000 2.872 1.971 .170
Lower-bound 2.872 1.000 2.872 1.971 .170
Error(TRIAL) Sphericity Assumed 45.178 31 1.457
Greenhouse-Geisser 45.178 31.000 1.457
Huynh-Feldt 45.178 31.000 1.457
Lower-bound 45.178 31.000 1.457
DAY * TRIAL Sphericity Assumed 5.686 1 5.686 4.719 .038
Greenhouse-Geisser 5.686 1.000 5.686 4.719 .038
Huynh-Feldt 5.686 1.000 5.686 4.719 .038
Lower-bound 5.686 1.000 5.686 4.719 .038
Error(DAY*TRIAL) Sphericity Assumed 37.351 31 1.205
Greenhouse-Geisser 37.351 31.000 1.205
Huynh-Feldt 37.351 31.000 1.205
Lower-bound 37.351 31.000 1.205

Tests of Between-Subjects Effects

Measure: MEASURE_1
Transformed Variable: Average
Type III Sum
Source of Squares df Mean Square F Sig.
Intercept 218728.237 1 218728.237 1867.847 .000
Error 3630.156 31 117.102
Table 2
Summary of the 2X2 repeated measures ANOVA (Day X Trial) for the test
data.
Source SS df MS F p
Subjects 3630.16 31 117.10
DAY 0.44 1 0.44 0.21 0.647
Error(DAY) 64.12 31 2.07
TRIAL 2.87 1 2.87 1.97 0.17
Error(TRIAL) 45.18 31 1.46
DAY * TRIAL 5.69 1 5.69 4.72 0.038
Residual 37.35 31 1.21
 One IV is between subjects, other IV is within
subjects
Analyze  General Linear Model  Repeated Measures
 Estimated Marginal Means of MEASURE_1
42.5

42.0

41.5
Estimated Marginal Means

41.0

SEX
40.5
Female

40.0 Male
1 2

TRIAL
 Tests of Within-Subjects Effects

Measure: MEASURE_1
Type III Sum
Source of Squares df Mean Square F Sig.
TRIAL Sphericity Assumed 8.144 1 8.144 4.773 .037
Greenhouse-Geisser 8.144 1.000 8.144 4.773 .037
Huynh-Feldt 8.144 1.000 8.144 4.773 .037
Lower-bound 8.144 1.000 8.144 4.773 .037
TRIAL * SEX Sphericity Assumed .159 1 .159 .093 .762
Greenhouse-Geisser .159 1.000 .159 .093 .762
Huynh-Feldt .159 1.000 .159 .093 .762
Lower-bound .159 1.000 .159 .093 .762
Error(TRIAL) Sphericity Assumed 51.193 30 1.706
Greenhouse-Geisser 51.193 30.000 1.706
Huynh-Feldt 51.193 30.000 1.706
Lower-bound 51.193 30.000 1.706

Tests of Between-Subjects Effects

Measure: MEASURE_1
Transformed Variable: Average
Type III Sum
Source of Squares df Mean Square F Sig.
Intercept 108429.072 1 108429.072 2027.326 .000
SEX 23.134 1 23.134 .433 .516
Error 1604.514 30 53.484
Table 3.
Analysis of the 2X2 ANOVA (Gender x Trial) with repeated
measures on the second factor for the test data.
Source SS df MS F p
SEX 23.13 1 23.13 0.43 0.516
Error 1604.51 30 53.48
TRIAL 8.14 1 8.14 4.77 0.037
TRIAL * SEX 0.16 1 0.16 0.09 0.762
Residual 51.19 30 1.71
 Table 1
Results of the 2-way ANOVA (Gender X Education) for Income.
Source SS df MS F p
Gender 12,142,860,574 1 12,142,860,574 28.95 0.000
Education 28,840,936,475 5 5,768,187,295 13.75 0.000
Gender X Education 6,215,428,787 5 1,243,085,757 2.96 0.012
Residual 197,550,969,717 471 419,428,810

Table 2
Summary of the 2X2 repeated measures ANOVA (Day X Trial) for the test
data.
Source SS df MS F p
Subjects 3630.16 31 117.10
DAY 0.44 1 0.44 0.21 0.647
Error(DAY) 64.12 31 2.07
TRIAL 2.87 1 2.87 1.97 0.17
Error(TRIAL) 45.18 31 1.46
DAY * TRIAL 5.69 1 5.69 4.72 0.038
Residual 37.35 31 1.21

Table 3.
Analysis of the 2X2 ANOVA (Gender x Trial) with repeated
measures on the second factor for the test data.
Source SS df MS F p
SEX 23.13 1 23.13 0.43 0.516
Error 1604.51 30 53.48
TRIAL 8.14 1 8.14 4.77 0.037
TRIAL * SEX 0.16 1 0.16 0.09 0.762
Residual 51.19 30 1.71
 HYPOTHESES TESTED
in 2-WAY ANOVA

 No differences for IV #1 (A - 3 levels)

These are
 H0: MA1 = MA2 = MA3
called
 No differences for IV #2 (B - 4 levels) “Main
Effects”
 H0: MB1 = MB2 = MB3 = MB4
 No interaction
 At least one MAiBj  MAmBn