The reason I ask is that SPSS tell me I have a quasi-complete separation of my data and after some reading I have found that using penalized maximum likelihood could help me solve this problem.
Is there other ways (maybe simpler) that could help me solve the problem?
I have two treatments and one co-variate (affect the response variable but is also correlated with one of the treatments). Each replicate has been examined for the presence/absence of morphological characters at several time point but there are unfortunately a lot of missing data. What would be...