Page 1 of 1
Monte Carlo simulation for panel data
Posted: Wed Feb 19, 2014 1:04 pm
by basher
I want to conduct a Monte Carlo simulation to decide whether a pooled OLS or a random effects model is ideal for my (unbalanced) panel data. Any help is highly appreciated.
Re: Monte Carlo simulation for panel data
Posted: Wed Feb 19, 2014 1:58 pm
by TomDoan
What type of MC simulation would answer that?
Re: Monte Carlo simulation for panel data
Posted: Wed Feb 19, 2014 2:13 pm
by basher
I guess I framed the question incorrectly. Using a Monte Carlo simulation can I decide whether a pooled OLS or a random effects model is appropriate for my data? Is it still an inappropriate question?
Re: Monte Carlo simulation for panel data
Posted: Thu Feb 20, 2014 10:33 am
by TomDoan
Do you have a reference? I have no idea what you're thinking of doing.
Re: Monte Carlo simulation for panel data
Posted: Fri Feb 21, 2014 11:22 am
by basher
I do not have a reference. I have an unbalance panel. I want to check the finite sample properties of random effects (RE), fixed effects (FE) and pooled OLS for my data (N=25, avg. T=10). I applied the Hausman test and found that RE is chosen over FE. Between RE and pooled OLS models, the Breusch-Pagan LM test supports RE. But all these diagnostic checks have certain assumptions (e.g., required a balanced panel), which may not be satisfied by my data. So I thought that conducting a small-sample simulation for these estimators based on my own data instead of simulated data would provide complementary evidence regarding the robustness of the RE model. I hope I am clearer this time.