BookmarkSubscribeRSS Feed
Highlighted
Srishti

Occasional Contributor

Joined:

Sep 12, 2018

matched pairs- conditional logistic regression using Choice Model platform

HI, 

 

I am a beginner in JMP (not familiar with JSL, would prefer an interactive explanation). I have data for a case-control study with 67 matched pairs. Each case is indivicually matched to a control (1:1). I need to use Conditional logistic regression to find out whether a particular outcome (column name: 'primary') is significantly different in cases vs controls, after controlling for several factors (column names: 'primary, arrest, distress, multifactor,presentation, CPD, AMA) . 

 

I tried running the analysis using the 'Choice model' platform in Consumer research. My data have been reformatted to what is required by this platform (as given in the example using Endometrial Cancer data library- file:///C:/Program%20Files%20(x86)/SAS/JMPPRO/13/Help/index.html#page/jmp%2FCR_Choice_19.html%23). I tried to run the analysis as shown in the figures attached, but am getting an error (shown in figure). I have highlighted one of the rows mentioned in the error report.

 

Any idea how to overcome the error and do the analysis? 

 

Thank you very much in advance, any help is greatly appreciated.

1 ACCEPTED SOLUTION

Accepted Solutions
cwillden

Community Trekker

Joined:

May 1, 2017

Solution

Re: matched pairs- conditional logistic regression using Choice Model platform

looks like the response indicator should be the Case/Control column, not primary.  Primary should be one of the factors I guess?  At least that's what it looks like with the example where the objective was to predict the chances of gall bladder disease based on the presence of endometrial cancer and controlling for hypertension.  Outside my wheelhouse, so if that's not right, you're out of luck with me.

-- Cameron Willden
2 REPLIES
cwillden

Community Trekker

Joined:

May 1, 2017

Solution

Re: matched pairs- conditional logistic regression using Choice Model platform

looks like the response indicator should be the Case/Control column, not primary.  Primary should be one of the factors I guess?  At least that's what it looks like with the example where the objective was to predict the chances of gall bladder disease based on the presence of endometrial cancer and controlling for hypertension.  Outside my wheelhouse, so if that's not right, you're out of luck with me.

-- Cameron Willden
Srishti

Occasional Contributor

Joined:

Sep 12, 2018

Re: matched pairs- conditional logistic regression using Choice Model platform

Yes, it looks like that was the mistake there. Thanks a lot!