Complex ANCOVA

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash


Complex ANCOVA

Postby MrEagle » Fri Apr 04, 2008 10:07 am

Hello there everyone, I'm incredibly stuck and could use some advice from the mathematically skilled.

I've conducted an entirely repeated-measures experiment, and now need to control for a co-variate. My problem is that every book I have consulted, and website, only give guidance for between-participants techniques.

I know that a repeated measures ANCOVA can be carried out. If anybody knows how to carry out this repeated measures procedure, and what tests of assumptions are required (you get a different output for RM so I'm uncertain how you would test for homogeneity of regression slopes - if you even need to in RM). What part of the output is relevant too?

Thanks for reading even if you don't know the answer. As I say I'm utterly stumped on this one.

Posts: 310
Joined: Wed Jun 27, 2007 8:58 pm
Location: Belgium

Postby yvesvv » Fri Apr 04, 2008 11:45 am

analyze - general linear model - repeated measures.

You can add the covariate in the covariates-box;)
Thank you for the nice comments;)
Posts: 2
Joined: Fri Apr 04, 2008 12:07 pm

Re: Complex ANCOVA

Postby MrEagle » Fri Apr 04, 2008 12:20 pm

Thanks for your reply yvesvv. Unfortunately I already only know how to get that far. I need advice about the possible tests of assumptions in RM-ANCOVA and some pointers on basic interpretation of the output, specifically related to the ANCOVA. RM-ANCOVA appears somewhat more ambiguous than the simpler between participants method.
Posts: 2757
Joined: Tue Jun 12, 2007 12:08 pm
Location: Florida, USA

Postby statman » Fri Apr 04, 2008 12:33 pm

What version of SPSS?

Have you looked into the Case Studies, generally they are very helpful especially since around V14

You might also ask specific questions about the output tables + content/results so we can visualize what you are seeing
See the note below

NOTE: Please read the Posting Guidelines and always tell us your OS, the SPSS version and information about your study and data!

Statistical Services
Posts: 310
Joined: Wed Jun 27, 2007 8:58 pm
Location: Belgium

Postby yvesvv » Fri Apr 04, 2008 1:04 pm

I agree with statman. Plus, there are multiple good textbooks which deal with repeated measures, maybe you can find one in a library.

It's not that simple to provide the whole explanation, so you might find the basics over there.
Thank you for the nice comments;)
Posts: 2
Joined: Fri Apr 04, 2008 12:07 pm

Postby MrEagle » Fri Apr 04, 2008 1:15 pm

Hi, thanks for your replies. I've looked at several higher textbooks, including ones written by people such as Andy Field an SPSS wizzard and they don't cover RM-ANCOVA, believe me, I wish they did!

I'm using SPSS version 14.

Basically I've run the repeated measures analysis on one IV with 3 levels, and a single covariate. IV is therapy type, DV is a depression score (so each participants data is along a single row as it should be - this is all corret and i'm happy here), covariate is an anxiety score.

Although this might sound a little dumb, I'm confused at the output table:

'Tests of Within-Subjects Effects'

Therapy type - sphericity assumed, sig 000. fair enough!
Therapy type*anxiety score (the IV*covariate) - spher assumed, sig.378

Is therapytype*anxiety score an interaction? Or is it the main effect minus the influence of the covariate?

I suppose I'm asking what the actual output of this analysis means in terms of the relationship with the covariate.
In a between-participants table of effects you are given 'intercept' 'iv' and 'covariate' to work it all out - nice and easy. Whereas all you get in a RM is this IV*covariate line.

My apologies if this question is frustrating to those skilled in this confusing business - i'm only a struggling yet enthusiastic MSc student.

Much thanks for your comments.
Posts: 1
Joined: Tue Dec 03, 2013 6:46 pm

Re: Complex ANCOVA

Postby dantepicchioni » Tue Dec 03, 2013 7:23 pm

Dear Colleagues,

Hello everyone. I am trying to conduct a three-way, fully repeated measures ANCOVA with SPSS 19 running on Mac OS 10.8.4. My dependent variable is continuous, my three independent variables are discrete/categorical, and my covariate is continuous and time-invariant/constant.

I would like to revive this post because it left off exactly where I need some help. I also though it would be a good idea to revive this post because an earlier post (Post subject: interpreting a three way interaction? Posted: Sat Mar 22, 2008 8:24 am) also left off exactly at this point.

When you have a fully repeated measures ANOVA and you enter your covariate in the covariate portion of the Repeated Measures GLM dialog box in the SPSS graphical user interface, your Tests of Within Subjects Effects table has two rows for every effect. For example, for the three-way interaction, I have Var1 * Var2 * Var3 and Var1 * Var2 * Var3 * covariate. In other words, for every main effect and interaction, I have Effect/Interaction and Effect/Interaction * covariate.

After much searching, I found this article on the IBM support website: << ... wg21477023>>. This was very helpful because it told me the following. "If you only run one model that includes the covariates, the covariates appear in the WS portion of the model as interactions with the WS factor. This is not the same as partialling the covariates from the WS factor; it is what it looks like --a set of interaction terms." However, it did not tell me what to do if I actually wanted to partial the covariates from the WS factor(s) or if that is even possible.


Dante Picchioni

Who is online

Users browsing this forum: No registered users and 2 guests