Hi All,
I have panel data on public disclosures of working (or non-working) of new technology.
The data is an unbalanced panel of 10 years, with the primary dependent variable being a binary variable, i.e., working (1) / non-working (0).
I am interested in modelling the working of technology on technology-specific variables. However, I only observe the machine status (working/non-working) when the firm files for the disclosure.
I want to correct for sample selection bias by controlling the effect of non-disclosure in the second stage.
I am new to Stata and sample selection models. However, I have found two viable options:
1. Run xtprobit in the first stage determining disclosure and non-disclosure. Generate Inverse Mills Ratio (IMR) from this stage and control it in the second stage of probit regression.
Can we run a probit model with random effects in the second stage?
2. Run xtheckman command. (However, this method takes too long to implement, and I'm not sure about its applicability.)
Are these methods appropriate to address self-selection in panel data with binary dependent variables?
Thank you in advance.
Regards,
Hariom Arora
I have panel data on public disclosures of working (or non-working) of new technology.
The data is an unbalanced panel of 10 years, with the primary dependent variable being a binary variable, i.e., working (1) / non-working (0).
I am interested in modelling the working of technology on technology-specific variables. However, I only observe the machine status (working/non-working) when the firm files for the disclosure.
I want to correct for sample selection bias by controlling the effect of non-disclosure in the second stage.
I am new to Stata and sample selection models. However, I have found two viable options:
1. Run xtprobit in the first stage determining disclosure and non-disclosure. Generate Inverse Mills Ratio (IMR) from this stage and control it in the second stage of probit regression.
Can we run a probit model with random effects in the second stage?
2. Run xtheckman command. (However, this method takes too long to implement, and I'm not sure about its applicability.)
Are these methods appropriate to address self-selection in panel data with binary dependent variables?
Thank you in advance.
Regards,
Hariom Arora
