Rolling estimates with confidence intervals (graph)

[River Huang]

Dear all, I found this question here. Please ssc install rangestat. The question is, after running rolloing regressions such as:

webuse grunfeld, clear
rangestat (reg) invest mvalue, interval(year -6 0) by(company)
*drop if reg_nobs < 7

how can one obtain the time series plots of estimates along with 95% confidence intervals like

Any suggestions? Thanks.

[Nick Cox]

You would have to construct confidence intervals from the ingredients provided. For this example, serial dependence bites twice over to invalidate conventional procedures.

[River Huang]

Dear Nick, I see your point. At this moment, suppose that I generate the (as your concern, incorrect) confidence intervals by

gen high = b_mvalue+2*se_mvalue
gen low = b_mvalue-2*se_mvalue

How can I draw a graph like the one I provided above in Stata? Thanks.