Another logistic regression queston

[dinsya]

Hi,
I am analyzing data with polytomous outcome, and the predictor is in rate ratio.
after run the logistic regression procedure, the result looks weird so I wonder if I am runing the right program.
My data is looks like this:

factor1 factor2 are low, middle and high
cy cx cd are rate such as 3.45 2.25 5.12 6.45 etc

my program is
proc logistic data=try;
CLASS factor1/ param=ref ref=first order=internal;
MODEL cy(ref=first) =factor1/nodummyprint link=glogit;
RUN;

am i right? because the result will be very long since it will appear for each rate. Can I use rate as a predictor?

Could somebody help me please? I am totally confuse.

Thanks in advance

[Link]

You coding does look weird. What are you trying to model (in your own words)?

[dinsya]

Hi Link,
1. How do factor 1 & factor 2 influence the risk of cy cx cd? Is there any specific effect or pattern?

2. To what extent do factor1 and factor 2 explain the relationship between factor3 and cy cx cd?

You are absolutely right that my program is weird.
I try to fixed it like this, but not sure if it is right:

proc logistic data=try;
CLASS factor1/ param=ref ref=first order=internal;
MODEL cy=factor1;
RUN;

Would you help me please?
Thanks a lot

[Link]

Hmmm...though I've used proc logistic too many times to keep count of, I've never used it for multinomial outcomes. For that, I usually use proc genmod.

1. How do factor 1 & factor 2 influence the risk of cy cx cd? Is there any specific effect or pattern?

From the way you described this, I would set up three models. Factors 1 and 2 would be the independent variable in each. cy, cx, and cd would be the dependent variables in each of the models.

2. To what extent do factor1 and factor 2 explain the relationship between factor3 and cy cx cd?

I'm a little unsure of what you're asking here. Are you asking how much of the variance between factor3 and the dependent variables are explained by factor1 and factor2? Together, or independently? How would you use the results?

Try looking at this site: http://www.ats.ucla.edu/stat/sas/out...git_output.htm

It'll show you what you need to do.

PS. Sorry for the delay in replying. As you might have guessed, I'm quite busy.