- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hi All,
What statistics analysis is recommended if you have dichotomous independent variables and continuous depend variable? The depend variable values includes 0, 1 and values between 0 and 1 (the probability of an accident).
Thank you!
Thanks,
Jade
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
Your dependent (=outcome) variable is not entirely continuous.
As it is a proportion, it is bounded by 0 and 1 (included). So you have a limited dependent variable.
You have to look at the distribution of the outcome.
Based on that you can decide to fit a GLMM with Poisson distribution (PROC GLIMMIX).
But the safest choice is probably a GLMM with beta regression (also with PROC GLIMMIX).
You might have to tackle some non-convergence problems, but we can look into this at a later stage (should you stumble upon non-convergence).
Good luck,
Koen
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thank you!
If I choose another Binary dependent variable (0 &1 only, the independent variables are all dichotomous ), will logistic regression be recommended?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
http://support.sas.com/kb/24/188.html
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Since your response is a proportion, if you have the numerator and denominator values for each of the proportions, then you can use the events/trials response syntax in PROC LOGISTIC and fit a logistic model. See the example in the Getting Started section of the PROC LOGISTIC documentation. If those are not available or conceptually don't exist, they you can fit a fractional logistic model or a 4- or 5-parameter logistic model as described in this note.