I am trying to estimate an ordinal logistic regression with clustered standard errors on subject level using the MASS package's polr() function. (My dependent variable is ordinal with three factors, treatment is a dummy, randinterval is integer, period is integer, male is a dummy) Standard errors are clustered at the level of the individual subject via SubjectID. I already saw a similar thread here (R: Clustering standard errors in MASS::polr()), however I have noticed that the output only includes SE for the coefficients but not for the cut-off points from the ordered probit regression. So far, my code looks like this:
oprobit <- polr(factor(frequency) ~ treatment + randinterval + period + male ,
data=my_data, method = "probit", Hess = T)
summary(oprobit)
Call:
polr(formula = factor(frequency) ~ treatment + randinterval +
period + male, data = my_data, Hess = T,
method = "probit")
Coefficients:
Value Std. Error t value
treatment 0.05849 0.064516 0.9065
randinterval -0.05419 0.028686 -1.8890
period -0.01289 0.007973 -1.6163
male 0.01108 0.067770 0.1635
Intercepts:
Value Std. Error t value
0|1 -0.8934 0.1290 -6.9234
1|2 -0.3580 0.1278 -2.8009
Residual Deviance: 2686.905
AIC: 2698.90
--------------------------------------------------------------------------------
oprobit_subj_1 <- coeftest(oprobit, vcovCL, type='HC0', cluster=~SubjectID)
print(oprobit_subj_1, digits=3)
t test of coefficients:
Estimate Std. Error t value Pr(>|t|)
treatment 0.05849 0.14225 0.41 0.68
randinterval -0.05419 0.03428 -1.58 0.11
period -0.01289 0.00845 -1.53 0.13
male 0.01108 0.14561 0.08 0.94
To the best of my knowledge, STATA does include the cut-off points via the following function:
oprobit frequency treatment randinterval period male, vce(cluster SubjectID)
Is there a way to include the SE for the intercepts (= cut-off points (0|1, 1|2)) in R via MASS as well?
Thanks for your help in advance!
