Re: ANCOVA or two-way ANOVA

Posted by Bruce Weaver on Apr 25, 2012; 3:25pm
URL: http://spssx-discussion.165.s1.nabble.com/ANCOVA-or-two-way-ANOVA-tp5665074p5665176.html

I think you're getting hung up on terminology that arose in the days before everyone had a computer & stats package.*  In more up-to-date terminology, you have a general linear model with two dichotomous explanatory variables (Group and Sex).  Dichotomous variables can be treated as either categorical (fixed factors) or continuous (covariates)--you'll get the same results either way.  (It's convenient to use 0-1 coding if you treat them as continuous.)  To verify this, run your model via UNIANOVA with both variables as fixed factors, with both as covariates, and with one factor and one covariate.  Then run the model again using REGRESSION.  (For REGRESSION, you'll have to compute your own product term if you're including the interaction.)  

HTH.

Nico wrote

Hi all,

I'm looking at the differences in tooth size between two populations. My primary concern is tooth size but I also want to see if sex has an effect on the variable - or at least take the sex variable out of the equation.

Should I run an ANCOVA with sex as a covariate or run a two-way ANOVA with both tooth size and sex as the fixed factors?

Thanks!