Karl got an answer to this via official channels, but in case others are interested, you can get simple main effects tests in a couple of ways in SPSS Release 8.0. One is by using LMATRIX (and MMATRIX if necessary) to specify the appropriate linear combination(s) of model parameters to test. The other is to use the EMMEANS COMPARE option. If you specify

/EMMEANS(A*B) COMPARE(A)

you get an omnibus test for A at each level of B, plus all pairwise comparisons among the levels of A, again at each level of B. The pairwise comparisons are unadjusted (LSD) by default, but you can use the ADJ keyword on the EMMEANS subcommand to request Bonferroni or Sidak corrected comparisons and confidence intervals. If there are covariates, you can also specify the value(s) on the covariate(s) at which the means are to be computed (the mean of each covariate is the default). These options are available for any combination of between and within subjects factors. They can also be extended to simple, simple (or higher) main effects. For example:

/EMMEANS(A*B*C) COMPARE(A) gives tests for A at each level of B*C.

David Nichols Principal Support Statistician and Manager of Statistical Support SPSS Inc.

>---------- >From: lloyd[SMTP:lloyd+@PITT.EDU] >Sent: Friday, July 17, 1998 12:33 PM >To: SPSSX-L@UGA.CC.UGA.EDU >Subject: simple main effects in SPSS8.0 > >Hello - a quick question > >In SPSS8.0 - when one has a 2x2 anova or a 2x3 anova and an ordinal >interaction is found - how do you calculate simple main effects >involving the interaction? Is this done via contrasts in SPSS8.0 and if >so how is this accomplished. > >Thank you, > >Karl > > >------------------------------------- >Karl Brett Lloyd >University of Pittsburgh >Katz Graduate School >412-856-8571 >lloyd@vms.cis.pitt.edu >------------------------------------- >