Hi all,
I am estimating a bivariate tobit model using mvtobit in Stata. The command I used is
'mvtobit (y1=$x1) (y2= $x2 ), robust cluster(group)'.
where y1 and y2 are both censored at zero, and the regressors &x1 and $x2 are exogenous.
My data is unbalanced panel data. In each period, there are different groups, and each group contains two members. The idea of using mvtobit is to allow for the dependence of error terms for the two equations that estimate the behaviours of the two members. At the same time, it allows for different coefficients for the two members.
My questions are:
(1) Is it true that the 'robust cluster' options correct for the autocorrelation of the same member over time and the heteroskedasticity across groups in estimating the standard errors of coefficients?
(2) Are the 'robust cluster' options good enough for panel data estimation?
I would be very grateful for any suggestions you may give!
I am estimating a bivariate tobit model using mvtobit in Stata. The command I used is
'mvtobit (y1=$x1) (y2= $x2 ), robust cluster(group)'.
where y1 and y2 are both censored at zero, and the regressors &x1 and $x2 are exogenous.
My data is unbalanced panel data. In each period, there are different groups, and each group contains two members. The idea of using mvtobit is to allow for the dependence of error terms for the two equations that estimate the behaviours of the two members. At the same time, it allows for different coefficients for the two members.
My questions are:
(1) Is it true that the 'robust cluster' options correct for the autocorrelation of the same member over time and the heteroskedasticity across groups in estimating the standard errors of coefficients?
(2) Are the 'robust cluster' options good enough for panel data estimation?
I would be very grateful for any suggestions you may give!