I was doing instrumental variable estimation with panel data using random effects. In Stata I have used the command xtivreg(we have used random effects, because Hausman test rejected H0 hypothesis). But then I wanted to compare random effects with pooled OLS. Usually in Stata we use the command xttest0 with xtreg. But in our case we have xtivreg instead of xtreg and as the result xttest0 doesn't work (the error is: 'last estimates not found').
Is there a test for this situation at all? Which models does it compare. My assumption was that I can view IV with panel data as a panel regression in the transformed model (Projecting all variables on to the instrument space) and compare it to the pooled ols resulting from the same projection (equivalent to the IV on the pooled cross-setion data)
Another question is: how to use Breusch-Pagan LM test in Stata, when we have xtivreg?
Is there a test for this situation at all? Which models does it compare. My assumption was that I can view IV with panel data as a panel regression in the transformed model (Projecting all variables on to the instrument space) and compare it to the pooled ols resulting from the same projection (equivalent to the IV on the pooled cross-setion data)
Another question is: how to use Breusch-Pagan LM test in Stata, when we have xtivreg?
