Paul Norris posted on Friday, November 09, 2012 - 8:12 am
I am hoping to use the Bayes estimator to conduct an EFA of variables which contain counts (following a negative binomial distribution). For each case, in addition to the count measures of interest, I have a variable which measure the population at risk. In a regression model, I would enter the log of the population as an exposure variable (coefficient constrained to be equal to 1). Is it possible to include such exposure measures when conducting an EFA, or an I better calculating rates and then using these for the factor analysis?