[Maxence Morlet]

Thank you very much for the explanation. I'll give it a try.

[Comlanvi Martin Konou]

Dear David:
Thank you for Boottest. I am currently using it for panel data analysis. I run the code below:

xtpcse er insti dfii gdp etrk fdi inf remi remi2 time bm educ, hetonly corr(psar1)


boottest {dfii}

Results:
Wild bootstrap-t, null imposed, 999 replications, Wald test, Rademacher weights:
dfii

z = 2.6494
Prob>|z| = 0.0100

95% confidence set for null hypothesis expression: [.03417, .3271]



I do not understand the interpretation of the graph where p-values are in the y-axis and CI on x-axis. Is there any reading to help with the interpretation?

I have also attached the codes and the results.
Thank you advance

[David Roodman]

There is a whole paper about boottest: Roodman, D., J. MacKinnon, M. Nielsen, and M. Webb. 2019. Fast and wild: bootstrap inference in Stata using boottest. Stata Journal 19(1): 4-60.
boottest is not designed to run after xtpcse. It may just be treating your model as if the standard errors are classical, with no correction. The graph shows the p value as a function of the hypothesized coefficient value.

[Comlanvi Martin Konou]

Thank very much for your response. It is highly appreciated. I will go and get the paper.