I have a smallish data set with three randomised intervention groups ($n$s between 13 and 22), with pre-post variables. I want to run an ANCOVA with the pre-data as a covariate and post as an outcome to look for between-group differences. Each outcome gets its own univariate ANCOVA because the different predictor and outcome variables correlate. The issue is, depending on which variable I'm looking at, pretty much every assumption of normality and equality of variance etc. gets broken at some point. I THINK I can get around this issue by just using the ANCOVA part to control for the pre variables and then performing bootstrapping (using the bootstrap function on the ANCOVA in SPSS) and reporting bootstrap confidence intervals rather than F and P values.

I really need someone to tell me whether or not this is a valid approach though, because I've sort of had to piece together bits of information from various sources and I'm not confident enough to be certain.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.