* Adapted from Ordinal regression models by Williams & Quiroz
* https://methods.sagepub.com/Foundations/ordinal-regression-models
* Extended ologit using European Social Survey GB 2012

* Open the data, do descriptive stats, drop cases with missing
use https://www3.nd.edu/~rwilliam/statafiles/ESSGB2012Extract.dta, clear
fre *, tab(5)
keep if !missing(hmsfree, lifeworse, hincfel, feelecon, agedecade, female)

*** Problem 1 Ologit
* Constrained model -- demographic vars only
ologit hmsfree agedecade female, cformat(%9.3f) nolog
est store constrained
* Unconstrained model -- Attitudes + demographic vars
ologit hmsfree lifeworse hincfel feelecon  agedecade i.female, cformat(%9.3f)  nolog 
est store unconstrained
* Test to see if attitude vars improve the model
lrtest constrained unconstrained, stats
* Now get margins, adjusted predictions. margins & mtable 
* tell you pretty much the same thing but the output from mtable is 
* much easier to read.
margins female
mtable, at(female=(0 1)) dec(4)
margins, dydx(female)
mtable, dydx(female) dec(4)
* Test model assumptions
brant, detail

***  Problem 2 Mlogit
mlogit hmsfree lifeworse hincfel feelecon  agedecade i.female, cformat(%9.3f)  nolog b(1)
mlogtest, lr
mlogtest, lrcomb


* Problem 3 Shuttle data using Interval Regression
use https://www3.nd.edu/~rwilliam/statafiles/shuttle2.dta, clear
recode distress (1 = .)(2=1)(3=3), gen(dlow)
recode distress (1=0)(2=2)(3=.), gen(dhigh)
intreg dlow dhigh date temp, cformat(%9.4f) nolog
oprobit distress date temp, nolog cformat(%9.4f)