Negative Binomial Regression with Time Variable

[amyroberts]

I am running a regression in R and I am confused. I am trying to understand what factors affect debate participation in parliament and how this has changed over time. I want to run a regression in R, with a count variable as the dependent variable (the number of times each member has participated in debate in a specific policy area per year) and 4 explanatory variables - sex of the member (dummy variable), political party of the member (dummy variables for each of the 3 main parties), whether or not they hold a leadership position in their party (dummy variable) and the number of years since they were elected (categorical variable).

I want to run a regression analysis over 11 years (2000 - 2010) to understand if changes have taken place over time. I have another categorical variable for the years. I was going to do a negative binomial regression (as my dependent variable is a count variable), but I am not sure how to analyse how the debate participation has changed over time. I don't think I can just include my year variable as an explanatory variable? Should I do a two step regression? Any help is much appreciated.

 

[afos]

No need for two stages. You can include Time and Time^2 as extra predictors. You may also consider interactions between Time and the four predictors that you have mentioned.

However, it is not immediately clear that negative binomial regression is the best choice. You can try four frameworks: Poisson regression, Poisson regression + zero-inflated component, negative binomial regression and negative binomial regression + zero-inflated component. Within each framework keep only the statistically significant terms. At the end, choose the best framework using objective model selection tools.