I have a data set of two variables as class and test score. Where class is a classification made by experts and test score ranges 0-10. The data set looks like:

Class   score
1        5.61
1        4.23
2        6.78
3        8.34
2        7.42
3        9.59

How can I find the boundaries in test scores which match classification? I'm looking for something like,say,

"0-5.15= class 1; 5.16-7.81= class 2; 7.82-10= class 3"

  • $\begingroup$ would the -party- package version of CART work since it includes ordinal responses? $\endgroup$ – charles Nov 15 '13 at 2:37
  • $\begingroup$ @charles is it an R package? $\endgroup$ – Amin Nov 15 '13 at 4:33
  • $\begingroup$ yes. -rpart- and -party- are the two packages I've used for CART (cran.r-project.org/web/packages/party/vignettes/party.pdf). But this is just one possibility. Think there are a number of other possible approaches. Could likely even split dataset into (c1&C2)(c2&c3) and -rpart- package and maybe even use logistic. Would have to think that through $\endgroup$ – charles Nov 15 '13 at 14:18
  • $\begingroup$ @charles Thanks.I'll take a look at those packages. The way I came across with is using multinomial regression twice.First setting the last category as reference, by setting ln[P(Class=2)/P(Class=3)] equal to 0 I can find corresponding score for boundary between second and third category. Then by setting the second category as reference I can find boundary between the first and second category by setting ln[P(Class=1)/P(Class=2)] equal to 0. $\endgroup$ – Amin Nov 16 '13 at 5:31

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.