1
$\begingroup$

I am currently looking at continuous data and seeking a correlation with the presence (and severity, ie, normal, mild, moderate, severe) of a disease entity. Now, so far, running Spearman on SPSS, I've gotten an R of 0.370, p<0.001.

I need help on a couple of things. Can/Should I perform a multinomial regression analysis? Also, I'd like to generate a ROC and find out corresponding cut-off values that may pertain to the severity of the disease, i.e., Normal, Mild, Moderate, Severe.

Please advise on how to go about this. Any input is highly appreciated.

Improve this question
$\endgroup$
8
  • $\begingroup$ Welcome to the site. What is/are your dependent and independent variables (if any)? Regression analysis is for relating a DV to one or more IVs. Multinomial is appropriate if the DV is categorical with more than 2 categories that are not ordered. That doesn't sound like the case you are dealing with. Why are you trying to find cut-off values? Categorizing a continuous variable is rarely a good idea. $\endgroup$
    – Peter Flom
    Oct 30, 2012 at 10:55
  • $\begingroup$ Thanks for the reply, @Peter! The DV is the disease. It is subcategorized into its severity, ie 0-normal, 1-mild, 2-moderate, 3-severe. I am trying to find cut-off values because my hypothesis is the bigger my IV is (which is a unit of measurement), the more severe my DV category is. I'd like to arrive at something like <3 is normal, 4-5 is mild, 6-7 moderate, >7 severe. $\endgroup$
    – ome
    Oct 30, 2012 at 11:15
  • $\begingroup$ If your independent variable is continuous, there is no need to categorize it. You can do ordinal logistic regression with an ordinal DV (disease) and a continuous IV. $\endgroup$
    – Peter Flom
    Oct 30, 2012 at 11:28
  • $\begingroup$ I see. Now, can I proceed with ROC analysis? If yes, do I use the real values or do I use the predicted probabilities? I apologize, I'm really confused. $\endgroup$
    – ome
    Oct 30, 2012 at 11:59
  • $\begingroup$ You should be able to generate something at least similar to an ROC curve from those results. What software are you using? $\endgroup$
    – Peter Flom
    Oct 30, 2012 at 12:04

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.