Dear STATA users
I have searched the Internet for clues on how to ”handle” a regression model using Random Effects. However, I have not found what I was looking for. I found a lot of information about fixed effects models and fGLS models, but little information on Random effects modeling.
I have a dataset consisting of 50 individuals and 7 time periods.
First I ran a Hausman test
xtreg dependent variable independent variables, fe
estimates store fixed
xtreg dependent variable independent variables, re
estimates store random
Hausman
estimate fixed random
Which showed me, that I ought to use a random effects model. This makes sense, as the variable of interest do not change much doing the time period.
In order to use my regression estimates, I would like to test for heteroskedasticity and autocorrelation:
In order to test for heteroskedasticity in a fixed effects model, I can use the likelihood ratio test described by Wiggins & Poi (2001) (which should also work for an xtgls model). However, how do I check for heteroskedasticity in a random effects model (xtreg, re)?
I know I can use -xtserial- to estimate serial correlation in a fixed effects model and in a xtgls model. However, how is it possible to check for serial correlation in a random effects (xtreg, re) model?
I know how to correct for any of above circumstances, I am just a little unsure how to test for heteroskedasticity and autocorrelation in a random effects model. I hope someone has the time to help me. I will appreciate any help at all.
Thank you so much in advance
Kind regards
Mia Ravn
I have searched the Internet for clues on how to ”handle” a regression model using Random Effects. However, I have not found what I was looking for. I found a lot of information about fixed effects models and fGLS models, but little information on Random effects modeling.
I have a dataset consisting of 50 individuals and 7 time periods.
First I ran a Hausman test
xtreg dependent variable independent variables, fe
estimates store fixed
xtreg dependent variable independent variables, re
estimates store random
Hausman
estimate fixed random
Which showed me, that I ought to use a random effects model. This makes sense, as the variable of interest do not change much doing the time period.
In order to use my regression estimates, I would like to test for heteroskedasticity and autocorrelation:
In order to test for heteroskedasticity in a fixed effects model, I can use the likelihood ratio test described by Wiggins & Poi (2001) (which should also work for an xtgls model). However, how do I check for heteroskedasticity in a random effects model (xtreg, re)?
I know I can use -xtserial- to estimate serial correlation in a fixed effects model and in a xtgls model. However, how is it possible to check for serial correlation in a random effects (xtreg, re) model?
I know how to correct for any of above circumstances, I am just a little unsure how to test for heteroskedasticity and autocorrelation in a random effects model. I hope someone has the time to help me. I will appreciate any help at all.
Thank you so much in advance
Kind regards
Mia Ravn
Comment