I need to compare two independent groups on a dependent variable while controlling for a covariate. Usually I would do an ANCOVA, but the dependent variable is non-normal (significant Shapiro-Wilk test - is this the correct way to test this?).
What is the best way to proceed? I can't see a way of controlling for a covariate using non-parametric statistics in SPSS. Given that ANCOVA is relatively robust can I just use that? If so would bootstrapping help at all?
Thanks for your help and apologies if this is a daft question! Solutions which use SPSS would be particularly appreciated.
You may want to give the JASP software a try. It is free to download and very easy to use, and it includes the option to conduct a bayesian ANOVA.
https://jasp-stats.org/download/
Take a look as well here:
Quintana, D. S., & Williams, D. R. (2018). Bayesian alternatives for common null-hypothesis significance tests in psychiatry: a non-technical guide using JASP. BMC psychiatry, 18(1), 178.
Whether the regular ANCOVA is robust will depend on your sample size and how strongly the distribution deviates from normality. When the sample size larger than 100 or so and the distribution is not extremely zero-inflated or something, a regular ANCOVA should still be possible.
Hi Stefanos, Safa and Roderik,
Thanks for taking the time to respond to my question! I'll look into you various pieces of useful advice.
Fergus
In Finland a sample size > 30 is considered good enough for parametric measures.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS