turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- multivariable analysis procedure for ordinal, cont...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-23-2017 05:27 AM

My outcome variable is ordinal in nature (1.e. score- high, moderate, mild and low).

I have about 15 independent variables that are either continuous, binary or ordinal in nature. I utilized Mantel Hanzel for binary variables X ordinal response, Logistic model for ordinal variables X ordinal response for those whose proportional odds criterion was satisfied and Glogit for ordinal variables X ordinal response for those whose proportional odds criterion WASNT satisfied and Glogit for Continuous variable X ordinal response. I got Values for OR and p-value for the variables as mentioned above. Are there any other procedures that

1 Are there any other procedures that I can utilize for Ordinal Dependent Variable?

2 I reported the ORs for these values. Now I have to choose all the variables that are significant. That should be done on the basis of CI of the OR estimates or the p-value?

3 After choosing the variables that are significant, which procedure can be used to check for multivariable analysis? Will Proc Logistic will be helpful to analyze several independent variables with ordinal response variable?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Arushiarora0

07-26-2017 03:18 PM

Doing a bunch of models or table analyses will muddy the water. You can use model selection in PROC LOGISTIC to select variables and model effects for a suitable single model for your ordinal response. It can even decide whether to allow for unequal slopes or not for a particular predictor. See this note that discusses the proportional odds issue. And see in particular the "Using SELECTION= to choose a partial proportional odds model" section that shows how to use model selection. More details and more examples are given in Derr (2013).

BTW, this note shows all the various types of logistic models and the procedures that can fit them.