"Equation not found" - Hurdle model using suest

[Jessica Weaver]

Hi there,
I am using Stata 15.

I am trying to estimate the marginal effects for a logit-truncated negative binomial hurdle model. Unfortunately I do not have the citation because I am using photocopies of pages from a book given to me by my advisor. The author of the book estimates the two parts of the model and then uses suest followed by a series of local macros followed by margins to estimate the marginal effects. I am following his method and it seems I am running into a silly error that I can't figure out.

Here's my code:

quietly svy linearized, subpop(SPD):logit useanyhcru i.yearind i.sex i.raceeth2 i.region i.married i.education ib1.natstat c.pcs ib1.agecat ib1.povcat2 i.insstat1
estimates store anyh1
quietly svy linearized, subpop(SPD):tnbreg anyhcru i.yearind i.sex i.raceeth2 i. region i.married i.education ib1.natstat c.pcs ib1.agecat ib1.povcat2 i.insstat1 if anyhcru>0, ll(0)
estimates store anyh2
quietly suest anyh1 anyh2

/*hurdle nb model marginal effects*/
local logit "invlogit(predict(eq(anyh1_useanyhcru))) "
local ey "exp(predict(eq(anyh2_anyhcru))) "
local pygt0 "(nbinomialtail(exp(-predict(eq(/anyh2:lnalpha))),1,""1/(1+exp(predict(eq(anyh2_anyHCRU)))/exp(-predict(eq(/anyh2:lnalpha))))))"
margins, dydx(*) expression("`logit'*`ey'/`pygt0'")

where "anyhcru" is the number of health care visits and "useanyhcru" is the number of health care visits conditional on having at least one visit (useanyhcru>0).

When I run the model, I get the following error:

. /*hurdle nb model suest*/
. quietly svy linearized, subpop(SPD):logit useanyhcru i.yearind i.sex i.raceeth2 i.region i.married i.education ib1.natstat c.pcs ib1.agecat ib1.povcat2 i.insstat1

. estimates store anyh1

. quietly svy linearized, subpop(SPD):tnbreg anyhcru i.yearind i.sex i.raceeth2 i. region i.married i.education ib1.natstat c.pcs ib1.agecat ib1.povcat2 i.insstat1 if anyhcru>0, ll(0)

. estimates store anyh2

. quietly suest anyh1 anyh2

. /*hurdle nb model marginal effects*/
. local logit "invlogit(predict(eq(anyh1_useanyhcru))) "

. local ey "exp(predict(eq(anyh2_anyhcru))) "

. local pygt0 "(nbinomialtail(exp(-predict(eq(/anyh2:lnalpha))),1,""1/(1+exp(predict(eq(anyh2_anyHCRU)))/exp(-predict(eq(/anyh2:lnalpha))))))"

. margins, dydx(*) expression("`logit'*`ey'/`pygt0'")
Warning: cannot perform check for estimable functions.
equation anyh2_anyHCRU not found
r(303);

I don't understand why it can seem to find the equation anyh1_useanyhcru but it can't find the equation anyh2_anyhcru. Any thoughts?

[Jessica Weaver]

**EDIT: ' "useanyhcru" is the number of health care visits conditional on having at least one visit (anyhcru>0).'

[William Lisowski]

I don't understand why ... it can't find the equation anyh2_anyhcru.

What it tells you it cannot find is the equation anyh2_anyHCRU (upper or lower case makes a difference).