Proc GLM and MACRO

[CarolBarahona]

Hi all! 

I have a confussion to perform GLM with a macro. 

I have this: 

A	B1	B2	B3	B4
0.1	0	2	1	1
0.9	1	1	0	2
0.78	2	1	1	1
0.64	0	2	0	0

and I need to perform several regression over variable the A. For example: A ~B1, A~B2, A~B3 etc.... (I have ~300 of variables). 

Later, I need all the p-values of each association to evaluate if one of the associations is true. 

But I don't have clue how to do it... 

I really appreciate your help, as always!

Thanks in advance!

[PaigeMiller]

All explained here: https://blogs.sas.com/content/iml/2017/02/13/run-1000-regressions.html

No macros needed.

Also, in my opinion, not a good idea to perform regressions like this, hundreds or thousands of variables individually.

[CarolBarahona]

Maybe, I didn't explain very well, but my variable X is categorical (just 3 values are possible) and Y variable is continuous.
So, I think is not possible to perform that code, or am I wrong?

[PaigeMiller]

Use PROC GLM instead of PROC REG.

[CarolBarahona]

Yes, but I don't know how to modify the previous code that you send me... sorry! I tried to applied to my dataset and I don't have results...

[PaigeMiller]

Show us the code you tried.

[CarolBarahona]

I'm trying with something simple like this:

proc glm data=exp;
class b1;
model a=b1;
run;

but I need to incorporate a macro because I have B1-B300 and also modified to retain just the p-values for each association.

[PaigeMiller]

You don't need a macro. The link I gave explains how to do this without a macro. Scroll down to the section entitled "The BY way for many models". Give that a try. If you get stuck, show us your code.

[CarolBarahona]

I modified the code but the problem now, is that the code don't give the p-value for each association evaluated. What is coded as "value" in proc print is not p-value of the association. 

data try1;
set exp; /* <== specify data set name HERE */
array new_SNP [*] new_SNP1 - new_SNP280; /* <== specify explanatory variables HERE */
do varNum = 1 to dim(new_SNP);
VarName = vname(new_SNP[varNum]); /* variable name in char var */
Value = new_SNP[varNum]; /* value for each variable for each obs */
output;
end;
drop new_SNP:;
run;

/* 2. Sort by BY-group variable */
proc sort data=try1; by VarName; run;

/* 3. Call PROC REG and use BY statement to compute all regressions */
proc glm data=try1 noprint outest=PE;
by VarName;
model south_ref = Value;
quit;

/* Look at the results */
proc print data=PE(obs=5);
var VarName Intercept Value;
run;

[PaigeMiller · Accepted solution]

I think you need to add a CLASS statement to your PROC GLM, since your variable named value is CLASS and not continuous.

If you want the overall p-value for the model, you also need the ODS OUTPUT statement like this:

ods output overallanova=pe;
proc glm data=try1 noprint;
by VarName;
class value;
model south_ref = Value;
quit;