I'd like to compute a facet model with two dimensions: form (direct and indirect aggression) and function (reactive and proactive). Each item is assigned to one characteristic on each dimension, such that, e.g., "I threaten people to get what I want" would be both direct and proactive. Here's a syntax of the model:
direct BY agg01 agg02 agg03 agg04; indirect BY agg05 agg06 agg07 agg08; react BY agg01 agg03 agg05 agg07; proact BY agg02 agg04 agg06 agg08;
form BY direct indirect; function BY react proact;
form WITH function@0; ! not sure if this makes sense ! b/c items belong to both dimensions
Now, one form and one function factor have negative residual variances which are too big to fix to zero. What else could I do (and does the model make sense at all)?