Leverage plots for fixed effects Poisson model?

[Mansi Jain]

Hi all,

I'm trying to figure out the influence of a few outlier observations on my results. I believe the way to do that with a linear model is to use leverage plots, and I'm presuming the same holds for an FE Poisson model. I can't figure out how to implement those though, and am also wondering if there are other ways to look at the influence of outliers. Would love some help figuring that out!

Thank you very much!
Mansi

[Mike Lacy]

I'd think instead of examining how much slope estimates change when each observation is left out (or, better yet, all observations corresponding to each value of the panel variable are left out), in spirit similar to what -dfbeta- would do. Per a thread on StataList several months ago about something like this, one can do this with the -jackknife- command, saving the results to a file. I have not done this in quite your situation, but I believe something as simple as this might be close:

jackknife _b, cluster(YourPanelVariable) saving("YourOutputFile.dta") : xtpoisson .........

(Corrections from someone knowledgeable about the niceities of doing this in the panel context would be welcome.)