Regression with maximum value for the output variable? (Clinical score)

Question

I am trying to regression an output variable which has a set maximum value (say 100). But when I try to predict it (using both linear and non-linear regression), the predicted values are sometimes greater than 100.

Is there a way to constraint that my predictions are less than or equal to 100 (other than the simple idea of converting the predicted values greater than 100 to 100?).

What is your outcome variable and why should it not exceed 100? This will determine the most appropriate approach. — Frans Rodenburg, Aug 17 '20 at 23:02
My output variable is a clinical score. But for argument sake, it can even be the total marks in a final exam where the maximum marks is 100. — Ed Wati, Aug 18 '20 at 09:17
Consider using beta regression after converting to a 0-1 range. See eg https://eeecon.uibk.ac.at/~zeileis/papers/ERCIM-2010.pdf — BigBendRegion, Aug 18 '20 at 12:09
@BigBendRegion, Thanks a lot. Will give beta regression a go!! — Ed Wati, Aug 18 '20 at 18:32

score 0 · Answer 1 · answered Oct 20 '21 at 14:55

0

Your outcome variable, a clinical score, seems to be an ordinal variable. So try ordinal regression. See Ordinal regression vs. summated score or Analysis for ordinal categorical outcome, and search this site.

answered Oct 20 '21 at 14:55

kjetil b halvorsen

63,378
26
142
467

Regression with maximum value for the output variable? (Clinical score)

1 Answers1