0

I am running a regression model to identify the relationship between exposure to hate speech and the adoption of hate speech on a prominent forum, controlling for a host of other variables. Theoretically, I feel it makes sense for my measure of hate speech exposure to be lagged by t-1 time periods, to capture how exposure the previous month influences behaviour the following month.

The panel dataset I am using is an unbalanced one and contains around 12,000 forum members, whose posting behaviour is analysed over a period of 63 months.

Is there anything I should be weary of when implementing a fixed effects regression model in these circumstances?

Connor95
  • 11
  • Yes: https://stats.stackexchange.com/questions/196578/difference-of-dynamic-panel-nickell-bias-and-the-incidental-parameter-probl/196674#196674 See https://en.wikipedia.org/wiki/Arellano%E2%80%93Bond_estimator for a potential solution. – Christoph Hanck Jul 11 '23 at 14:34
  • thanks Christoph - though that appears to be relevant for a lagged Y variable on the right hand side of the regression equation. In my case, I am using a lagged X. – Connor95 Jul 12 '23 at 10:17
  • Oh, in this case I misread your question. In that case there is less reason for concern – Christoph Hanck Jul 12 '23 at 13:25
  • Many thanks for the clarification! I was hoping that was the case. – Connor95 Jul 12 '23 at 15:14

0 Answers0