ANCOVA with robust standard error estimates in SPSS: syntax to obtain main effect?

paola

New Member
#1
I am editing my post from March 4 because I have been doing some more research and I understand from the forum guidelines it is not considered inappropriate to change your question if you have not received any answers yet.
I also think it would be better to move this post to the SPSS forum but I am unable to do so (I have looked at the general forum Q&A but it seems to me you can't do it as a simple poster).
So please forgive me if I am doing something wrong!

I am using SPSS version 25 to run an ANCOVA with heteroskedastic-consistent standard errors estimators (HC3 procedure) and I am puzzled by some of the output I get.

I am fine with the robust standard errors estimates table with the significance levels for the comparisons of the dependent variable across my three groups (which are different from those of the "standard" ANCOVA as I would expect them to be).

What I really can't understand is where I can find the main effect of group for my dependent variable since in the output (between subjects effect table) the F statistic and its significance are exactly the same as those of the "standard" ANCOVA.

It would seem to me that the main effect statistics should also be different. I have been trying to run a linear regression with the same data, both a "standard" linear regression and a robust linear regression using Hayes' macro (afhayes.com/spss-sas-and-mplus-macros-and-code.html) and the F value and the p value for the overall model fit are different as well as the p-values for the coefficients (although of course the coefficient themselves are not).

Also, I found the following sentence in Andy Field's new SPSS book (p 620):"...you can also ask for CI and p values for parameter estimates that are robust to heteroscedasticity. This won't help for the F statistics".

I don't know what to do because I need to first report the main effect of group on my dependent variable and then compare it across groups.

So, my questions are:

1) Is it an SPSS issue ?
2) If yes, is there any SPSS syntax I could use to overcome the problem?

I hope someone can help. Thank you so much in advance for any tips/suggestions (or for pointing out some conceptual mistake I am unaware of....)


PS: In case more context is needed: I am using ANCOVA to compare three patient groups (fairly balanced n1=78,n2=82, n3=93) across a dependent variable (heart rate) adjusting for three independent variables (hypertension, depression score and anxiety score). Since the assumption of homoskedasticity is not satisfied I decided to go for a robust ANCOVA. From the drop-down menu I selected Analyze--> general linear models--> univariate and ultimately ticked the HC3 box.
 
Last edited: