I am currently using the nominal_test function, in the ordinal package, to test for the proportional odds assumption but have encountered a problem.

Here is my code:

feelingsmodel<-clm(Feelings ~ 
                     Gender + 
                     Location + 
                     Role + 
                     FarmType + 
                     Age + 
                     Income + 
                   data = data, na.action=na.exclude, Hess = TRUE , link = "logit")


And my output:

Tests of nominal effects

formula: Feelings ~ Gender + Location + Role + FarmType + Age + Income + Knowledge
          Df  logLik    AIC     LRT Pr(>Chi)  
<none>       -629.12 1320.2                   
Gender     5 -625.92 1323.8  6.3829  0.27073  
Role      15 -627.18 1346.3  3.8806  0.99810  
Age       15 -618.83 1329.7 20.5712  0.15110  
Income    10 -617.82 1317.6 22.5870  0.01238 *


For some explanatory variables (Location, FarmType and Knowledge) there seems to be no output. When I run similar models different output for different explanatory variables is omitted seemingly at random. How I could solve this problem and get a p value for all explanatory variables?

  • $\begingroup$ Welcome to the site, @Jack Shepherd-Cross. What is your sample size? Do you have any residual degrees of freedom? If you do not, that would mean that you had insufficient degrees of freedom to fit coefficients for your explanatory variables, which could be the reason for the problem. Check the output for 'clm' as well, and make sure you have 'warnings' messages turned on, as they may give you some hints as to where any problems might lie. $\endgroup$ – Izy Jun 5 at 10:57
  • $\begingroup$ @Izy Thanks for your reply. My sample size is 603 so I doubt sample size is the problem. Warning messages are turned on and clm output seems to be ok! All seems a but strange. Thanks! $\endgroup$ – Jack Shepherd-Cross Jun 5 at 13:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.