I have a couple of requests for this already-excellent package.
Could you add the ability to interact the time FE with group FE? That is, suppose there is a unit-time panel and each unit exists within a larger group, and the researcher wants to use unit FE and time-group FE, as in $y_{igt} = \alphai + \gamma{gt} + \beta D{igt} + \epsilon{igt}$. E.g., there is a county-year panel, and you want to use county and state-year FE. The did2s package based on Gardner (2022) allows for this, so it should be feasible within the FEct framework?
Could you add the ability to use the FEct estimator on panels in which each row is a dyad-year observation and the researcher wants to use three layers of FE (a so-called "triple differences" design)? E.g., suppose we have a candidate-donor-cycle panel, and the researcher wants to estimate candidate-donor, candidate-cycle, and donor-cycle FE, as in $y{ijt} = \alpha{ij} + \gamma{it} + \delta{jt} + \beta D{ijt} + \epsilon{ijt}$. Strezhnev (2023) claims to use the Liu, Wang, and Xu (2024) imputed counterfactual method in this setting, so this should also be feasible?
I have a couple of requests for this already-excellent package.
Could you add the ability to interact the time FE with group FE? That is, suppose there is a unit-time panel and each unit exists within a larger group, and the researcher wants to use unit FE and time-group FE, as in $y_{igt} = \alphai + \gamma{gt} + \beta D{igt} + \epsilon{igt}$. E.g., there is a county-year panel, and you want to use county and state-year FE. The
did2spackage based on Gardner (2022) allows for this, so it should be feasible within the FEct framework?Could you add the ability to use the FEct estimator on panels in which each row is a dyad-year observation and the researcher wants to use three layers of FE (a so-called "triple differences" design)? E.g., suppose we have a candidate-donor-cycle panel, and the researcher wants to estimate candidate-donor, candidate-cycle, and donor-cycle FE, as in $y{ijt} = \alpha{ij} + \gamma{it} + \delta{jt} + \beta D{ijt} + \epsilon{ijt}$. Strezhnev (2023) claims to use the Liu, Wang, and Xu (2024) imputed counterfactual method in this setting, so this should also be feasible?
Thank you for the excellent package!