Discussion: View Thread

Dear All,

I am working on a panel dataset, which has 6-7 years of data for several countries, and the dependent variable is count while independent variables are mostly factor level. It's a terrorism related data and the subject matter expert would like to incorporate reverse causality.Usually the way this is dealt with in terrorism studies is to lag the dependent variable by one year (so that predictor values in say 1981 correspond to terrorist attacks in 1982). Does anyone  know how to deal with the issue, if one is using NB model, or mixed effect GLM in R.

thanks

Tasneem 

format of the data looks like below

country   Year     Y   X1  X2

A           1981     2  .........

A            1982    0 ....... 

------------------------------
[Tasneem] [Zaihra]
[Assistant Professor]
[SUNY-Brockport]
------------------------------

lagged variables are more common in econometrics , there's likely to be an R library for the estimation  

https://eml.berkeley.edu/wp/tsp_ref/negbin.pdf

http://www2.gsu.edu/~ecosgg/research/pdf/gp_ee.pdf

------------------------------
Chris Barker, Ph.D.
Consultant and
Adjunct Associate Professor of Biostatistics


---
"In composition you have all the time you want to decide what to say in 15 seconds, in improvisation you have 15 seconds."
-Steve Lacy