Hi, Here's my situation.
-Stata version: MP 13.1
-Data: Unbalanced panel data with time gaps
-Endogeneity: Yes, thus used instrumental variables
-Key independent variable: Time-invariant
-Time dummies are included in models.
Because the independent variable which is core in analysis is time-invariant, I've decided to use random effect (RE) model rather than fixed effect (FE) model. I have tried a Hausman test to be sure, but it didn't work when I ran -xtivreg-s. Thus, I just ran -xtreg-s with no IVs, then took a Hausman test which resulted in that H0 was not rejected.
Q1. Is this result of Hausman test (H0 not rejected) also valid when I use -xtivreg-, or do I need another result of Hausman test for -xtivreg-? If I need another result, how can I make that result?
Suppose that using RE model rather than FE model for -xtivreg- is confirmed. I want to select a model between -xtivreg, re- and -ivregress 2sls-. As far as I know, the latter is a kind of pooled OLS (of course, it is definitely not OLS), not of an RE model. Thus, I have tried to run a Breusch-Pagan LM test for RE (-xttest0-) after running -xtivreg, re-, but the test didn't work, displaying "last estimates not found." Unfortunately, I have not been able to find any solution for this problem.
Q2. Do you have any idea/criteria/method to select a model between -xtivreg, re- and -ivregress 2sls-?
Thanks in advance, and if you need information more in detail, please don't hesitate to tell me.
-Stata version: MP 13.1
-Data: Unbalanced panel data with time gaps
-Endogeneity: Yes, thus used instrumental variables
-Key independent variable: Time-invariant
-Time dummies are included in models.
Because the independent variable which is core in analysis is time-invariant, I've decided to use random effect (RE) model rather than fixed effect (FE) model. I have tried a Hausman test to be sure, but it didn't work when I ran -xtivreg-s. Thus, I just ran -xtreg-s with no IVs, then took a Hausman test which resulted in that H0 was not rejected.
Q1. Is this result of Hausman test (H0 not rejected) also valid when I use -xtivreg-, or do I need another result of Hausman test for -xtivreg-? If I need another result, how can I make that result?
Suppose that using RE model rather than FE model for -xtivreg- is confirmed. I want to select a model between -xtivreg, re- and -ivregress 2sls-. As far as I know, the latter is a kind of pooled OLS (of course, it is definitely not OLS), not of an RE model. Thus, I have tried to run a Breusch-Pagan LM test for RE (-xttest0-) after running -xtivreg, re-, but the test didn't work, displaying "last estimates not found." Unfortunately, I have not been able to find any solution for this problem.
Q2. Do you have any idea/criteria/method to select a model between -xtivreg, re- and -ivregress 2sls-?
Thanks in advance, and if you need information more in detail, please don't hesitate to tell me.
