which is dependent variable (proc GENMOD)

[RyanD]

I'm trying to determine the prevalence of risk factors in a Black population compared to a reference group, but I can't figure out if the risk factor is the dependent or independent variable.

For example (Here's my 2X2 table):
____________Risk Factor
____________Yes__________No
-------------------------------------------------------
Black________34_________99,966_______100,000
RefPop_______26_________99,974_______100,000
--------------------------------------------------------
Total_________60________199,940_______200,000

I've been calculating the RR as (34/100,000)/(26/100,000)=1.3

I interpret this to mean that the Black population is 1.3 times as likely to smoke (one of the risk factors) as the reference population.

However, when I test using proc GENMOD (NOTE: next step is to include covariates, but I wanted to test it to see if the calculated RR using a single risk factor matched my 2X2 table above) I can't get it to calculate the RR I got above (1.3). However, it does calculate to the alternative RR (34/60)/(99,966/199,940)=1.13

Here's my SAS code:

proc genmod data=dataset1;
class RiskFactor (ref='1') /param=ref;
model RefPop = RiskFactor / link=log dist=bin;
estimate 'Beta RiskFactor ' RiskFactor 1 -1/ exp;
run;

(Note: RefPop = 0 indicates Black, RiskFactor=1 indicates risk factor present)

This seems like a really fundamental concept that I can't wrap my head around in this case. What am I doing wrong?

To take this one step further, I've planned to set up the multivariate analysis like this:

proc genmod data=dataset1;
class RiskFactor1 (ref='1') RiskFactor2 (ref='2') RiskFactor3 (ref='1') /param=ref;
model RefPop = RiskFactor1 RiskFactor2 RiskFactor3 / link=log dist=bin;
estimate 'Beta RiskFactor1 ' RiskFactor1 1 -1/ exp;
estimate 'Beta RiskFactor2 ' RiskFactor2 1 -1/ exp;
estimate 'Beta RiskFactor3 ' RiskFactor3 1 -1/ exp;
run;

Is this correct?

Thanks for your help.

[Dale]

Ryan,

You do, indeed, have your independent and dependent variables turned around in your genmod code. Your regression model would suggest that the risk factor can affect whether a respondent is black or not. This is a strange notion indeed!

You can get the RR of 1.3 with the following code:

proc genmod data=test desc;
  class RefPop (ref='1') /param=glm;
  model RiskFactor = RefPop / link=log dist=bin;
  estimate 'Beta RefPop' RefPop 1 -1/ exp;
run;

[RyanD]

Thanks Dale. The code you provided worked! I appreciate your help.

[RyanD]

Hi Dale,

I'm getting an error message "Out of memory" when I run my code with the dist=poisson option. Do you know what's causing this or any work arounds?

Thanks,
Ryan