I am trying to reproduce in stata a similar model of Blinder and Bluhm (2010) (you can find online the working paper: http://www.cesifo-group.de/DocDL/cesifo1_wp3161.pdf ).
They develop a panel model with sample selection and endogeneity.
In page 8, they define their basic equations:
\[ y^*_{it}= \mu_i + d_{it} \theta +x'_{it} \beta + e_{it} , \] ("participation effects equation") (1),
\[ \]
\[ y_{it} = y^*_{it} \ if \ d_{it}>0 \]and "unspecified" otherwise.
\[ \]
\[ d^*_{it} = a_i + z'_{it} \gamma +u_{it} ,\] ("participation effects equation") (2),
\[ \]
\[ d^*_{it} = d^*_{it} \ if \ d^*_{it}>0 \] and \[0\] otherwise.
\[ x_{it} \ is \ a \ subset \ of \ z_{it}\]
They suggest a tobit fe estimation of (2) and panel fe for (1).
My first question is if they use a heckman tobit or just a tobit for estimating equation (2).
Then, then I would like to suggest me about the commands I should use, (e.g. for equation (1) an xtreg with fe is ok , or I have to use something else...)
Thanks a lot!