- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
I want to check if there is an interaction between sex and quartiles of soda consumption in the case of a binary outcome.
How should I proceed?
exposure: quartiles of soda consumption
outcome: presence or absence of depression
covariate sexe: female, male
Thank you in advance!
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Fit a logistic regression model including the interaction. These statements will include a test for the interaction (sex*consumption). It assumes that your response variable is a character variable ("present" or "absent"), that the event of interest is presence of depression, and that you want to treat consumption as a continuous effect in the model.
proc logistic;
class sex / param=glm;
model outcome(event="present") = sex consumption sex*consumption;
run;
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thank you very much!
In the case of a multinomial outcome, i.e.: different mental disorders, would the procedure be the same?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Logistic Regression ?
proc logistic data=have;
class sex;
model outcome=SEX QUARTILES SEX*QUARTILES;
run;
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thank you very much!
In the case of a multinomial outcome i.e.: different mental disorders, would the procedure be the same?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Yes, but if the response is nominal (unordered), you need to add the LINK=GLOGIT option in the MODEL statement. If it is ordinal (ordered), no change is needed.