Here are the same data again, together with the SAS datastep to input the data into a SAS data set named factor01. We then have 4 SAS PROC MIXED analyses, fitting the same fixed effects components (a, b and a*b), but with different combinations of variance components.
Here is the output from the first PROC MIXED run, where we have fitted a model with the main effects factors a and b as well as the interaction a*b with one common residual error variance for all observations.
Factor A has 2 levels (with values 1 and 2). Factor B has 3 levels (with values 1, 2 and 3). The estimate of the (co)variance, the residual error variance is 7.889.
The SAS System
The MIXED Procedure Class Level Information Class Levels Values
A 2 1 2 B 3 1 2 3
Covariance Parameter Estimates (REML) Cov Parm Estimate
Residual 7.88888889
Model Fitting Information for WT Description Value
Observations 24.0000 Res Log Likelihood -48.2889 Akaike's Information Criterion -49.2889 Schwarz's Bayesian Criterion -49.7341 -2 Res Log Likelihood 96.5777
Tests of Fixed Effects Source NDF DDF Type III F Pr > F
A 1 18 0.53 0.4767 B 2 18 46.18 0.0001 A*B 2 18 7.93 0.0034
The SAS System
The MIXED Procedure Class Level Information Class Levels Values
A 2 1 2 B 3 1 2 3
REML Estimation Iteration History Iteration Evaluations Objective Criterion
0 1 63.49596156 1 1 60.19416416 0.00000000 Convergence criteria met.
Covariance Parameter Estimates (REML) Cov Parm Group Estimate
DIAG A*B 1 1 2.25000000 DIAG A*B 1 2 13.33333333 DIAG A*B 1 3 12.91666667 DIAG A*B 2 1 3.58333333 DIAG A*B 2 2 8.25000000 DIAG A*B 2 3 7.00000000
Model Fitting Information for WT Description Value
Observations 24.0000 Res Log Likelihood -46.6380 Akaike's Information Criterion -52.6380 Schwarz's Bayesian Criterion -55.3091 -2 Res Log Likelihood 93.2760 Null Model LRT Chi-Square 3.3018 Null Model LRT DF 5.0000 Null Model LRT P-Value 0.6536
Tests of Fixed Effects Source NDF DDF Type III F Pr > F
A 1 18 0.53 0.4767 B 2 18 56.92 0.0001
The SAS System
Tests of Fixed Effects Source NDF DDF Type III F Pr > F
A*B 2 18 7.73 0.0038
The SAS System
The MIXED Procedure Class Level Information Class Levels Values
A 2 1 2 B 3 1 2 3
REML Estimation Iteration History Iteration Evaluations Objective Criterion
0 1 63.49596156 1 1 60.80351835 0.00000000 Convergence criteria met.
Covariance Parameter Estimates (REML) Cov Parm Group Estimate
DIAG B 1 2.91666667 DIAG B 2 10.79166667 DIAG B 3 9.95833333
Model Fitting Information for WT Description Value
Observations 24.0000 Res Log Likelihood -46.9427 Akaike's Information Criterion -49.9427 Schwarz's Bayesian Criterion -51.2782 -2 Res Log Likelihood 93.8853 Null Model LRT Chi-Square 2.6924 Null Model LRT DF 2.0000 Null Model LRT P-Value 0.2602
Tests of Fixed Effects Source NDF DDF Type III F Pr > F
A 1 18 0.53 0.4767 B 2 18 56.92 0.0001 A*B 2 18 7.73 0.0038
The SAS System
The MIXED Procedure Class Level Information Class Levels Values
A 2 1 2 B 3 1 2 3
REML Estimation Iteration History Iteration Evaluations Objective Criterion
0 1 63.49596156 1 1 63.11253690 0.00000000 Convergence criteria met.
Covariance Parameter Estimates (REML) Cov Parm Group Estimate
DIAG A 1 9.50000000 DIAG A 2 6.27777778
Model Fitting Information for WT Description Value
Observations 24.0000 Res Log Likelihood -48.0972 Akaike's Information Criterion -50.0972 Schwarz's Bayesian Criterion -50.9875 -2 Res Log Likelihood 96.1943 Null Model LRT Chi-Square 0.3834 Null Model LRT DF 1.0000 Null Model LRT P-Value 0.5358
Tests of Fixed Effects Source NDF DDF Type III F Pr > F
A 1 18 0.53 0.4767 B 2 18 46.18 0.0001 A*B 2 18 7.93 0.0034