I am looking to assess the predictive validity of two measures (M1 and M2) on a binary outcome. After constructing my model, do I just calculate odds ratios and compare the odds ratios for the two measures?
$\begingroup$
$\endgroup$
5
-
$\begingroup$ What do you mean by predictive validity? There are several options: AUC, Brier-score, Sensitivity vs specifitity plot, etc. $\endgroup$– AnalystCommented Jun 17, 2018 at 19:10
-
$\begingroup$ I'm trying to see if M1 or M2 predicts the outcome better. I though AUC was to assess the validity of the model rather than specific variables within the model $\endgroup$– iressa13Commented Jun 17, 2018 at 19:28
-
$\begingroup$ Yes, it is global metric. So you want to compare if other variable is more powerful than other in some sense? What are their measurement units, that affects things? $\endgroup$– AnalystCommented Jun 17, 2018 at 19:30
-
$\begingroup$ yes that is what I want to do. The measurement units differ with one being categorical (0, 1, 2) and the other being count (1-27). $\endgroup$– iressa13Commented Jun 17, 2018 at 20:23
-
$\begingroup$ I think it will be quite different to have a change from category value 0 to 1 than to move one unit upwards in scale of your continuous variable... $\endgroup$– AnalystCommented Jun 18, 2018 at 19:26
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Thanks for the additional info.
Now I can only suggest you to use likelihood ratio based test where change in the LR statistic without other variable is the interesting outcome.
