Testing the Homogeneity of Slopes in Repeated measures ANOVA SPSS

Question

I conducted a repeated measures ANOVA in SPSS, using the command Generalised Linear Model > Repeated Measures. I have a within subject factor with two levels (A), a measure (B) and a covariate (C). The within subject variables based on the factor are AB_1, AB_2.

According to the literature, ANCOVA has an additional assumption, the one of homogeneity of slopes. I have been trying to conduct the test in SPSS using the syntax commands I found here : https://www.ibm.com/support/pages/testing-homogeneity-slopes-hos-assumption-factorial-ancova-spss, which however refers to factorial ANCOVA. Specifically, I wrote:

DATASET ACTIVATE DataSet2.
GLM AB_1 AB_2 WITH C
  /WSFACTOR=A 2 Simple(1) 
  /MEASURE=B
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(B)
  /PRINT=DESCRIPTIVE HOMOGENEITY 
  /CRITERIA=ALPHA(.05)
  /WSDESIGN=B
  /DESIGN=C*A

I keep getting the following error:

A syntax error was detected in the effect specification where the symbol dur was encountered. Effect specification should have the following structure effect = BY-expression [WITHIN BY-expression [...]] BY-expression = variable [BY variable [...]] where variable is a factor or a covariate.

My question is if there's any was to test this assumption for repeated measures designs?

Answer 1

I'm not sure "dur" comes from, but there are a few things wrong in the syntax.

First, B is a measure name rather than a factor, and doesn't belong on WSDESIGN. You can remove the WSDESIGN subcommand altogether, or substitute A for B there.

Second, substitute A for B on the PLOT subcommand for the same reason.

Finally, A is a within-subjects effect, and can't go on the DESIGN subcommand, which specifies the between-subjects part of the model. You can remove the DESIGN subcommand altogether or just change C*A to C.

GLM uses the multivariate general linear model approach to repeated measures, with some post-estimation finagling to get some repeated-measures tests and such. In this model, all within-subjects factors are automatically crossed with each other, and also with anything from the between-subjects part of the design, which in this case is the covariate C. So you'll get an interaction of A and C automatically.