wild bootstrap on Stata 18 with multinomial logit

[Shisho Jakas]

Hi statalist,

I am trying to do wild cluster boostrap

wildbootstrap mlogit Y X

but the message says
"estimator mlogit not allowed; estimator must be one of regress, xtreg, or areg"

Is there any possible way to use wildbootstrap for multinomial logit?

[Rich Goldstein]

I believe that -boottest- will do this; see

h boottest

[Shisho Jakas]

Dear Rich,

Thank you for your suggestion. However, it seems like "boottest" does not provide wild clustering bootstrap for "mlogit" command. When I read the help file, it says the default of boottype for ML is score test.

I went for the boottest but also the message says :

(h0(1) assumed)
Constraint 1 not found and will be skipped.
conformability error

For your reference, the code i tried was:

eststo: mlogit sector i.agecohort##c.distance female, vce(cluster village)  base(1)
boottest, cluster(village)

the "sector" is a categorical variable for industry sector (4 categories).

To explain more about my model, I am struggling with small number of villages (cluster) so I would like to address it with the wild bootstrap by Cameron, Gelbach, and Miller (2008).

How can I further go with boottest?

Could you give me some suggestions for correction or any other code?

Thank you so much.