Bootstrapping with lots of dummies! Help!

[Cooper Felix]

Reviewer asked me to bootstrap standard errors, but I have 116 firm dummies plus 15 time dummies in addition to 14 control variables and the sample size is N=746. When I use

vce(boot)

or

bootstrap, reps(100)

Stata always told me that "insufficient observations to compute bootstrap standard errors". Can I assume that bootstrapping is not going to work with my sample given that there are too many fixed effects included? Any tips or suggestions? Many thanks.

-- A concerned author whose paper is likely to be rejected because of the issue with standard errors

[Carlo Lazzaro]

Cooper:
I'm not clear with the type of regression you performed.
I assume that you have a panel data but your data layout follows the -wide- format.
If that were the case, I would switch to the -long- format via -reshape-.
Stata throws a warning message that relate to your data, not the procedure: hence, inspect your dataset (eg, have you got a relevant share of missing values for some variables).
Eventually, the recommended number of bootstrap replications for SE ranges from 45 to 200 (https://www.routledge.com/An-Introdu...9780412042317; page 47), with the upper limit of the range being the usual number.

[Cooper Felix]

Thank you Carlo, this is very helpful! The following code works!

areg ... , a(firm ID) vce(bootstrap, reps(200))

Cooper:
I'm not clear with the type of regression you performed.
I assume that you have a panel data but your data layout follows the -wide- format.
If that were the case, I would switch to the -long- format via -reshape-.
Stata throws a warning message that relate to your data, not the procedure: hence, inspect your dataset (eg, have you got a relevant share of missing values for some variables).
Eventually, the recommended number of bootstrap replications for SE ranges from 45 to 200 (https://www.routledge.com/An-Introdu...9780412042317; page 47), with the upper limit of the range being the usual number.