I am trying to compute average treatment affect using endogenous switching regression model (i.e. Lokshin, Michael and Zurab Sajaia, 2004) to estimate the impact of soil conservation adoption on yield and income in Stata 14.2. The model runs fine., but I'm not sure how to proceed to compute the average treatment effects.
Please help me this regard with a command or by illustrating how to estimate the treatment effect? . Thank you in advance.
recently the movestay commend is appropriate if you see this paper it will be so clear https://www.stata-journal.com/sjpdf.html?articlenum=st0071 ... see page 284 for the full syntax of the command
the estimates can be generated using mspredict once you run the above syntax
I've been working on figuring this out myself recently. It seems the 2004 article output and post hoc prediction syntax is fairly different from what is needed now in 2019.
One change is that instead of labeling the output as regime 1 and regime 0, followed by rho1 and rho2 (further below in the output), which leads to some confusion about which rho pertains to which regime, the current output is regime 0 and regime 1, followed by rho0 and rho1, which is more obvious. This is important, but doesn't answer the original question.
To do average treatment effects, use the predict command (not mspredict). For example:
. predict ate0 if e(sample), yc0
. predict ate1 if e(sample), yc1
. summarize ate0 ate1
I chose ate0 and ate1 as the newvar names of the variables the predict command creates. The e(sample) option tells Stata to include only observations that were included in the movestay equations. Summarize is just a quick way to display the means of the fitted values.
Thank you for your recommendation. Such procedure may provide us to predict ate0 and ate1 but we couldn't be sure that whether the difference is statistically significant or not. I have seen some published works which reported the level of significance together with the average treatment effects. Please recommend me if there are some commands related to it.