10-25-2013 12:55 PM
I'm currently working with a dataset and created 4 quartiles using proc rank called education4. What I'm trying to do next is create indicator variables from those quartiles.
The quartiles according to proc rank are 0,1,2,3
Here is my code:
if education4=0 then ed1=1; else ed1=0;
if education4=1 then ed2=1; else ed2=0;
if education4=2 then ed3=1; else ed3=0;
if education4=3 then ed4=1; else ed4=0;
ed1 ed2 ed3 ed4 are obviously the indicator variables that i'm trying to create with ed1 being what would be known as the reference level. However, when I try to run proc logistic which only includes ed2 ed3 and ed4 i don't get an error message but rather something that says the following parameters have been set to 0, since the variables are a linear combination of other variables as shown.
SAS then tells me the parameter estimates for ed2 ed3 and ed4 are all 0 which isn't what i'm looking for.
Can anyone tell me what this all means? and how I can correct for this so I get the actual parameter estimates? Thanks!
10-25-2013 01:18 PM
Is that the exact code you submitted? You don't have to have data=, but I find it helpful for debugging especially.
Run a proc freq as follows and see what comes up, post if possible:
proc freq data=have;
Your proc logistic code should also be equivalent to
proc logistic data=have;
class education4/param=ref ref=first;
10-25-2013 01:39 PM
Then that's why SAS is giving you the estimates it is. There's something wrong somewhere then but with what you've provided we can't tell you.