I am having trouble using polr for a ordinal regression because I also have ordinal independent variables. Looking online it looks like I should either convert them to categorical dummy variables or go the numerical continuous route. Does anyone have any advice for which is the best option?
I have been able to get the model to run but R runs several polynomials for the independent ordinal variables (L, Q, C etc.). Is there a way to only have one run or is the best option to lose the order and make them continuous.
One model example is that the dependent variable is willingness to work (very unwilling, unwilling, neutral, willing, very willing) and one of the independent variables is level of work experience (none, some, moderate, significant).
Q132o <- factor(first_wave$Q132, levels=c("VU", "SU", "N", "SS", "VS"), ordered=TRUE) Q33o <- factor(first_wave$Q33, levels=c("N", "So", "Mo", "Si"), ordered=TRUE) MF6 <- polr(Q132o~ Q33o, data = first_wave, Hess = TRUE)
summary_MF6 <- coef(summary(MF6)) pvalMF6 <- pnorm(abs(summary_MF6[, "t value"]),lower.tail = FALSE)* 2 summary_MF6 <- cbind(summary_MF6, "p value" = round(pvalMF6,3)) summary_MF6 Value Std. Error t value p value Q33o.L 0.39314058 0.4363851 0.9009029 0.368 Q33o.Q -0.16563280 0.3518657 -0.4707274 0.638 Q33o.C -0.34313835 0.2858676 -1.2003399 0.230
Another issue is that when I don't have the factors ordered I am told that the model is "rank deficient and that coefficients are dropped"
