# Fitting generalized linear model with covariate

In our experimental design, we are assessing the effect of TMS on decision making. Our data is not normally distributed because our assesment test results in mean count data (ultimatum game). Besides, even randomized the previous test decisions are different across subjects as it is not possible to equalize in our case. I would prefer to rely on ANCOVA with pretest scores as covariate if they were normally distributed (not sure). But I decided to use a generalized linear model (GENLIN procedure in SPSS) instead, and use the binomial distribution with log link (as there is also too many zeros in the data set with a high variance) in SPSS.

My question is, can I use the pretest score as covariate, posttest as dependent and group as factors for GENLIN as we do it within the ANCOVA framework in many randomized trials in psychology?

• What you call GLZ is actually known as a GLM :) (fixed in my edits) Could you indicate what you mean by "equalize"; when you speak about "normally distributed" scores, do you refer to the response variable or the covariate? – chl Nov 15 '11 at 11:12
• General - izing the Linear Model The Generalized Linear Model is an extension of the General Linear Model to include response variables that follow any probability distribution in the exponential family of distributions. The exponential family includes such useful distributions as the Normal, Binomial, Poisson, Multinomial, Gamma, Negative Binomial, and others. – cumhur Nov 15 '11 at 13:10
• As seen I am talking about GLZ intead of GLM, so my qoestion is in pre-post test experimental designs, can we use pretest as covariate as we do in GLM analysisi also in GLZ analysis. :), so they are different ;) – cumhur Nov 15 '11 at 13:12
• userwww.sfsu.edu/~efc/classes/biol710/Glz/… – cumhur Nov 15 '11 at 13:13
• Ok you mean GLM as general linear model (aka linear model). Never seen GLZ, though. – chl Nov 15 '11 at 13:53