Hi,
I'm analysing my survey data to test whether A affects B using the ordinal logistic regression model in R ("polr" function). In my case, both A (the explanatory variable) and B (the response variable) are categorical. Each has 5 levels, which are "Never", "Rarely", "Sometimes", "Often", "Every time". Since the explanatory variable is ordinal, the model gives 4 t-values comparing two adjacent levels. However, I'm only interested in testing whether A affects B in general, rather than the difference between different levels. Could you please tell me if there is any way to get the overall effect of levels in ordinal logistic regression using R? (i.e., to get only one t-value for the ordinal explanatory variable) Additionally, how do I test the assumption about proportional odds, and get only one OR and CI in this case? Many thanks in advance!
The Anova function in the car package is the essential function for this purpose. Something like the following grammar should work, where "model" is your model.
library(car)
Anova(model)
For post-hoc tests, you can use the emmeans package. ( https://cran.r-project.org/web/packages/emmeans/vignettes/models.html ). I've never used polr, but I have used the ordinal package. Something like the following grammar should work, where "model" is your model, "Location" is your independent variable or interaction.
library(emmeans)
marginal = emmeans(model, ~ Location)
marginal
pairs(marginal, adjust="tukey")
library(multcomp)
cld(marginal)
I am sorry but if your DV is really categorical and NOT ordinal. , You probably should be using multinomial logistic regression not ordinal logistic regression. Please check the definitions of these scales of measurement against your data set and if needed ask a new question. David Booth
David Eugene Booth are you sure the DV is not ordinal? I think we can define an order on the choices as:
"Never" < "Rarely" < "Sometimes" < "Often" < "Every time"
David Eugene Booth; Babak Jamshidi Hi! Thanks so much for your reply. Sorry that I didn't make it clear. Both of the dependent and independent variables are ordinal: "Never" < "Rarely" < "Sometimes" < "Often" < "Every time". I've changed my description of this question :)
Thanks very much @Babak. Please count the number of times the question says categorical. How the hell did I know so I asked. Ok?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Software
- R