I have the following data set:

Row    Patient        Freq of Sickness     Sick # of times(2 yrs)    Symp 1   Symp 2  Sym 3  
 1        A               60 days                        4              1        0       1          
 2        B               90 days                        10             0        1       1  
 3        C               40 days                        5              0        0       6
 4        D               30 days                        0              1        0       0 
 5        E               80 days                        1              3        0       1
  • What kind of logistic regression should I use if I need to predict the probability for each patient to be sick in the next year ? Also, need to calculate the maximum likelihood and odds ratio of each symp which contributes to the sickness.

  • Should I consider Sick # of times(2 yrs) as ordinal or nominal ? I think Ordinal.

Presently, I am using PROC Logistic and PROC GENMOD in SAS to solve this problem. In GENMOD i used Cumulative logit and in PROC Logistic I have tried using both Ordered logit and un ordered Glogit. I got the best results when I used cumulative logit.

Any reference to study material would be highly appreciated.

Also is Logistic Regression the correct approach?

  • 1
    $\begingroup$ Similar study available here $\endgroup$ – Metrics Jul 5 '13 at 16:19
  • $\begingroup$ I don't see an outcome variable?? $\endgroup$ – ReliableResearch Jul 5 '13 at 21:11
  • $\begingroup$ The outcome variable can be " Sick # of times". That's another confusion I have, that whether to use this as an independent variable or response variable. Because Sick # of time in past 2 years can also be a very good indicator. $\endgroup$ – learnlearn10 Jul 8 '13 at 12:51
  • 2
    $\begingroup$ If number of times sick is the outcome, is logistic regression really the correct approach? $\endgroup$ – ReliableResearch Jul 11 '13 at 15:03
  • $\begingroup$ I think we can use logistic regression to check the probability if the patient will be sick again or not. I used Sick # of times as the dependent variable in PROC LOGISTIC and got good results. Another thing that I am trying to do is to take each case as a separate row. When we do that, we will have 4 rows for patient A and 10 for B respectively. Let's see how that works out. $\endgroup$ – learnlearn10 Jul 11 '13 at 21:37

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.