STATS MAKE ME CRY

Hi Jeremy,

Sorry for the double post (I posted something similar as a comment on your blog about conducting a Repeated Measures MANCOVAs in SPSS). I was hoping to provide you with more information than I initially did to hopefully provide you with enough information to get me un-stuck with my current results for a manuscript I am hoping to write up in the next few weeks.

I have determined that Repeated Measures MANCOVA is the most appropriate test for the questions I am hoping to answer.

Here is a summary of the the data I am using:

3 continuous DVs that were collected at two time points (T1 and T2)
2 between-subjects categorical IVs (1 has two levels and the other has 5 levels)
3 covariates that I need to account for (SES, Gender, Race)

I am not really interested in the covariates other than to make sure I am accounting for them but when I run the analyses in SPSS my output has ton of information on these variables, including main effects and interactions with my IVs, that I don't really want to have to report in a paper as it will distract from main findings regarding my between and within subject variables.

I am wondering if there is a way to account for the covariates but not include them in the analyses in SPSS, such as using residuals? If so, any help in making sure I am doing this correctly would be greatly appreciated.

I look forward to your reply!

Best,
Kristen

Thanks Jeremy! I completely agree that I would like to include them in the analysis (and would be the preferred option) but is it okay if I dummy-code gender and ethnicity to add them to the "covariates" section or do they have to go to in "Between Subjects" box? If I treat them as "Between Subject" variable the output provides me with a ton of interactions that I am not interested in- which is what I was really trying to avoid. I basically want to figure out a way to account for them without having to interpret their interactions with my IVs, much like I would do with step-wise regression where I enter the covariates first. My other question is whether it is okay to have a covariate that is related to both your IVs and DVs. Ideally we would want the covariate to only be correlated with the DV but unfortunately gender and ethnicity are related to everything in my study - including my IVs!
Thank you SOO much! I have been wracking my brain wondering if I am doing everything correctly.
Best,
Kristen

Also, if I do include dummy-coded gender and ethnicity as covariates in my MANCOVA are the marginal means even interpretable? Thanks!!!

Hi Jeremy,

Thanks so much for getting back to me. Glad to know that dummy coding and putting them in as covariates is an okay option. I actually tried the customizing option but it still didn't give me the option to label some of the variables as covariates and for some reason putting them in as Between-Subject variables (even just for main effects and no interactions) gives me a completely different results from any other time I have run my data (either with ANOVAs or Regression) or when I put them in as just covariates (but dummy coded). I wish I new more to understand why that might be the case.

Glad to know the marginal means would be not be interpretable - that is what I was thinking. So I am wondering how I go forward with interpreting significant 2-way interactions (X1= 2 levels, X2=5 Levels). Do you know of any SPSS code that would allow me to specific the contrasts for an ANCOVA model with three covariates? Or how I figure out where the significant differences are located if I can't use the marginal means or the typical option to select "Tukey" post-hoc is not available since I have covariates?

I am slowly figuring this out!!!

Thanks!