Hello Statalisters,
I am running an IV specification looking at the impact of social capital on the type of healthcare service used - public, private, alternative, none (similar to Hollard and Sene, 2016). However, I have been running this with binary outcome variables i.e. public = 1 if person reports using public healthcare and 0 otherwise, similarly for private and other healthcare options. I am using Sergio Correa's -ivreghdfe- which is modelled after -ivreg2-, so the code looks a bit like:
However, I feel like I am losing a lot of context by making dichotomous variables as opposed to using multinomial outcomes which would give me a distance to a reference and be better if I want to calculate marginal effects. How do I instead run -ivreghdfe- with multinomial outcomes? Thanks.
I am running an IV specification looking at the impact of social capital on the type of healthcare service used - public, private, alternative, none (similar to Hollard and Sene, 2016). However, I have been running this with binary outcome variables i.e. public = 1 if person reports using public healthcare and 0 otherwise, similarly for private and other healthcare options. I am using Sergio Correa's -ivreghdfe- which is modelled after -ivreg2-, so the code looks a bit like:
Code:
gen pub = 0
replace pub == 1 if hcs_type = 11 | hcs_type == 12 | hcs_type == 13 | hcs_type == 14
gen pvt = 0
replace pvt == 1 if hcs_type = 21 | hcs_type == 22 | hcs_type == 23 | hcs_type == 24
.
.
.
local all `r(varlist)'
local demog "i.religion educ i.res i.inccat i.sex"
local district "sexratio2011 literacy2011 hospnum"
foreach var in pub pvt alt none{
ivreghdfe `var' `demog' `district' (soccap = IV) closelect, clust(dist) partial(i.religion educ i.res i.inccat i.sex sexratio literacy hospnum) first
}
