fixed effects poisson model- mediation possible?

[JesMHill]

Hi

I have a panel data set consisting of 600+ individuals measured at 4 intervals. My outcome variable is a measure of delinquent acts, and is count, therefore I need to run a poisson model. My predictor is a continuous measure (0-10) of 'feeling adult'. This is the model I ran:

xtpoisson delinquency adult, fe (effect of feeling adult on delinquency is significant)

I want to test whether the negative effect of feeling adult on delinquency is mediated by how antisocial an individuals freiends are. This antisocial friends variable is also a count measure and thus heavily skewed.

So far I have run these models:

xtpoisson delinquency antisocial, fe (effect of antisocial friends on delinquent acts is significant)

xtpoisson antisocial adult, fe (effect of adult on antisocial friends is significant)

xtpoisson delinquency adult antisocial, fe (effect of adult is reduced)

It seems to me that partial statistical mediation is possible, but is there any way to test this with fixed effects poisson models?

Thanks so much for any help.

Jessica