Hi, I wonder, has anybody tried to visualize multinomial/ordinal with maginsplot, similarly as the one here (slide 24): https://www.ucl.ac.uk/population-hea.../files/lca.pdf but it works for binary manifest variables. I have categorical (ordinal with dif scales).
What I have tried is smth like this (thanks Chuck Huber for your assistance with those commands):
gsem (cuts police inequality global PARTY <- _cons), family(multinomial) link(logit) lclass(C 2)
estat lcprob
estat lcmean
margins, predict(outcome(1.cuts) class(1)) ///
predict(outcome(1.police) class(1)) ///
predict(outcome(1.inequality) class(1)) ///
predict(outcome(1.global) class(1)) ///
predict(outcome(2.PARTY) class(1))
marginsplot, recast(bar) xtitle("") ytitle("") ///
xlabel(1 "Cuts" 2 "Police" 3 "Inequality" 4 "Global" 5 "PARTY") ///
title("Predicted Probability of Behaviors For Class 1 with 95% CI")
… but it assumes specifying value for each DV. Given that in my case it would makes sense to treat my DVs as ordinal, I would prefer a version which is based on means (centered or standardized would also go) where I have all DV/manifest variable means across classes on one figure (like this one made by tidyLPA: https://rdrr.io/cran/tidyLPA/man/plot_profiles.html
Any ideas?
Thankfully,
0 Response to Marginsplot after multinomial LCA
Post a Comment