I am running a regression to test the impact of immigration on the UK native levels of employment. I made my data stationary using the second difference for all the variables. I have 4 variables, and using an OLS 3 of them are highly significant and an R-squared of about 96% (too high in my opinion. However is time series data (38 obs) so I do not know how much it matters ). Anyway, my data does not have serial correlation using the LM test (F statistic (0.547217), prob. F(2,31) - 0.5840). Also, it does not present heteroskedasticity when using a breusch pagan test. However, when using the white test I have heteroskedasticity (no Arch heteroskedasticity). My question is, in order to remove the problem of the general heteroskedasticity indicated by the white test should I use a robust OLS? I have tried to use it, my regressors remain significant and my R-squared dropped to about 56%. Is a robust OLS a more desirable test in my case?
For econometric discussions not necessarily related to EViews.
1 post • Page 1 of 1
Who is online
Users browsing this forum: Google [Bot] and 2 guests