# GLM - Time Series

I'm trying to fit a GLM-model based on negative binomial to a time series (I have data for the number of forrest fires in Sweden per every 3 months over a 20 year period) and wonder if I need to check for stationarity? If so, how would I do that?

I am using a package called "tscount" and my fitted model looks as follows: Y = $$\beta_0$$ + $$\beta_1Y(t-2)$$ + $$\beta_3$$Temperature_celsius + $$\beta_4$$Rain_mm + $$\beta_5$$Dummy(Vinter) + $$\beta_6$$Dummy(Spring) + $$\beta_7$$Dummy(Summer).

After checking the residuals I think the model looks okay, I have no autocorrelation etc.

Thanks!