an error occurred when svy executed mlogit: option baseoutcome() is required with prefix svy bootstrap

[Nader Mehri]

Hi,
Using the code below, I am trying to obtain and save marginal estimates for each iteration of my multinomial logisitic regression. I am getting the error: "an error occurred when svy executed mlogit: option baseoutcome() is required with prefix svy bootstrap". How can I fix this?
Thanks,
NM

set more off
use http://www.stata-press.com/data/r14/nhanes2f, clear
svyset psuid [pweight=finalwgt], strata(stratid) psu(psuid)

capture program drop savemargins
program savemargins, rclass
svy, subpop(race): mlogit health i.agegrp##c.zinc
margins, at(zinc=(0 10)) predict(outcome(1)) post
end

bootstrap _b, saving(margins, replace) reps(200): savemargins

[Joseph Coveney]

I am getting the error: "an error occurred when svy executed mlogit: option baseoutcome() is required with prefix svy bootstrap". How can I fix this?

Um, maybe by specifying a base outcome using the -baseoutcome()- option of -mlogit-?

See below.

.ÿ
.ÿversionÿ16.1

.ÿ
.ÿ/*ÿsetÿmoreÿoffÿ*/
.ÿ
.ÿclearÿ*

.ÿ
.ÿuseÿhttp://www.stata-press.com/data/r14/nhanes2f,ÿclear

.ÿ
.ÿsvysetÿpsuidÿ[pweight=finalwgt],ÿstrata(stratid)ÿpsu(psuid)

ÿÿÿÿÿÿpweight:ÿfinalwgt
ÿÿÿÿÿÿÿÿÿÿVCE:ÿlinearized
ÿÿSingleÿunit:ÿmissing
ÿÿÿÿÿStrataÿ1:ÿstratid
ÿÿÿÿÿÿÿÿÿSUÿ1:ÿpsuid
ÿÿÿÿÿÿÿÿFPCÿ1:ÿ<zero>

.ÿ
.ÿ/*ÿcaptureÿprogramÿdropÿsavemarginsÿ*/
.ÿprogramÿsavemargins,ÿrclass
ÿÿ1.ÿÿÿÿÿÿÿÿÿversionÿ16.1
ÿÿ2.ÿÿÿÿÿÿÿÿÿsyntax
ÿÿ3.ÿ
.ÿÿÿÿÿÿÿÿÿsvy,ÿsubpop(race):ÿmlogitÿhealthÿi.agegrp##c.zinc,ÿbaseoutcome(1)
ÿÿ4.ÿÿÿÿÿÿÿÿÿmargins,ÿat(zinc=(0ÿ10))ÿpredict(outcome(1))ÿpost
ÿÿ5.ÿend

.ÿ
.ÿtempfileÿmargins

.ÿbootstrapÿ_b,ÿsaving(`margins'ÿ/*,ÿreplaceÿ*/)ÿreps(20ÿ/*200ÿ*/):ÿsavemarginsÿ
(runningÿsavemarginsÿonÿestimationÿsample)

Bootstrapÿreplicationsÿ(20)
----+---ÿ1ÿ---+---ÿ2ÿ---+---ÿ3ÿ---+---ÿ4ÿ---+---ÿ5ÿ
....................

PredictiveÿmarginsÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNumberÿofÿobsÿÿÿÿÿ=ÿÿÿÿÿÿ9,188
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿReplicationsÿÿÿÿÿÿ=ÿÿÿÿÿÿÿÿÿ20

------------------------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿObservedÿÿÿBootstrapÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿNormal-based
ÿÿÿÿÿÿÿÿÿÿÿÿÿ|ÿÿÿÿÿÿCoef.ÿÿÿStd.ÿErr.ÿÿÿÿÿÿzÿÿÿÿP>|z|ÿÿÿÿÿ[95%ÿConf.ÿInterval]
-------------+----------------------------------------------------------------
ÿÿÿÿÿÿÿÿÿ_atÿ|
ÿÿÿÿÿÿÿÿÿÿ1ÿÿ|ÿÿÿ.1371592ÿÿÿ.0369608ÿÿÿÿÿ3.71ÿÿÿ0.000ÿÿÿÿÿ.0647175ÿÿÿÿÿ.209601
ÿÿÿÿÿÿÿÿÿÿ2ÿÿ|ÿÿÿ.1233056ÿÿÿ.0330288ÿÿÿÿÿ3.73ÿÿÿ0.000ÿÿÿÿÿ.0585704ÿÿÿÿ.1880408
------------------------------------------------------------------------------

.ÿ
.ÿuseÿ`margins',ÿclear
(bootstrap:ÿsavemargins)

.ÿ
.ÿlistÿinÿ1/2,ÿnoobs

ÿÿ+---------------------+
ÿÿ|ÿÿÿÿ_bs_1ÿÿÿÿÿÿ_bs_2ÿ|
ÿÿ|---------------------|
ÿÿ|ÿ.1016489ÿÿÿÿ.092148ÿ|
ÿÿ|ÿ.1981432ÿÿÿ.1738499ÿ|
ÿÿ+---------------------+

.ÿ
.ÿexit

endÿofÿdo-file


.

[Richard Williams]

I’ve never seen the error before but it makes sense. By default, mlogit chooses the category with the highest N as the base category. When bootstrapping, the category with the highest N could change from one bootstrap to the next.

Personally, I always choose the base category myself anyway. If I analyzed different sub samples I wouldn’t want the base category to keep changing.

[Joseph Coveney]

Personally, I always choose the base category myself anyway. If I analyzed different sub samples I wouldn’t want the base category to keep changing.

Good advice, and I will make it a point to do so in the future if I haven't always in the past.