topic Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro? in SAS Programming

How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Denali — Mon, 20 Nov 2017 20:28:54 GMT

Hi,

I wrote a macro for univariate logistic regression analysis. How do I export the odds ratio, 95% C.I. and p-value from each of the below model? Should I add ods output statement in the macro? Thank you!

%macro CDR_diff(x=);
proc logistic data=cdr_df1;
class &x;
where visitnum=1;
model cdr_dfcat2= &x / expb;
run;
%mend CDR_diff;
%CDR_diff(x=sex);
%CDR_diff(x=ethnic1);
%CDR_diff(x=race1);
%CDR_diff(x=smoke2);
%CDR_diff(x=alcohol1);
%CDR_diff(x=marital1);
%CDR_diff(x=educ1);
%CDR_diff(x=employ1);
%CDR_diff(x=educ_yrs1);
%CDR_diff(x=income1);

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

PeterClemmensen — Mon, 20 Nov 2017 20:39:03 GMT

Use ODS TRACE on before your PROC LOGISTIC and see what output you want the underlying data from and use ODS OUTPUT on the relevant output.

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Reeza — Mon, 20 Nov 2017 20:39:05 GMT

Should I add ods output statement in the macro?

Yes, and you should look at the PERSIST option within that.

Here's a very related question that was recently posted:

https://communities.sas.com/t5/SAS-Statistical-Procedures/multiple-logistic-models-output-odds-ratios/m-p/413512/highlight/true#M21685

@Denali wrote:

Hi,

I wrote a macro for univariate logistic regression analysis. How do I export the odds ratio, 95% C.I. and p-value from each of the below model? Should I add ods output statement in the macro? Thank you!

%macro CDR_diff(x=);
proc logistic data=cdr_df1;
class &x;
where visitnum=1;
model cdr_dfcat2= &x / expb;
run;
%mend CDR_diff;
%CDR_diff(x=sex);
%CDR_diff(x=ethnic1);
%CDR_diff(x=race1);
%CDR_diff(x=smoke2);
%CDR_diff(x=alcohol1);
%CDR_diff(x=marital1);
%CDR_diff(x=educ1);
%CDR_diff(x=employ1);
%CDR_diff(x=educ_yrs1);
%CDR_diff(x=income1);

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Denali — Mon, 20 Nov 2017 21:18:18 GMT

Hi,

I successfully output the odds ratio and confidence interval for the individual model using the "ods output OddsRatios(persist)=Oddsratio ModelInfo(persist)=mf;" statement.

But everytime I ran another predictor variable or another model, the odds ratio in the out was replaced. How do I keep all the odds ratios in one output file? For example, I need to run 10 models. How do I keep all 10 odds ratios and confidence intervals in the oddsratio file?

ods trace on;
%macro CDR_diff(x=);
ods output OddsRatios(persist)=Oddsratio ModelInfo(persist)=mf;
proc logistic data=cdr_df1;
class &x;
where visitnum=1;
model cdr_dfcat2= &x / expb;
run;
ods output close;
%mend CDR_diff;
%CDR_diff(x=sex);
%CDR_diff(x=ethnic1);
%CDR_diff(x=race1);
%CDR_diff(x=smoke2);
%CDR_diff(x=alcohol1);
%CDR_diff(x=marital1);
%CDR_diff(x=educ1);
%CDR_diff(x=employ1);
%CDR_diff(x=educ_yrs1);
%CDR_diff(x=income1);

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Reeza — Mon, 20 Nov 2017 21:21:29 GMT

Review the options for PERSIST.

And, have you considered transposing your data and just using a BY with a single PROC LOGISTIC call? Seems more efficient to me.

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Denali — Mon, 20 Nov 2017 21:41:42 GMT

Hi,

I tried the transpose statement but it did not work. Could you please check my code? Thank you!

%macro CDR_diff(y=);

ods output OddsRatios(persist)=ORCI ModelInfo(persist)=mf;

proc logistic data=cdr_df1;
class &y;
where visitnum=12;
model cdr_dfcat2= &y / expb;
run;

proc transpose data=ORCI out=Odds;
by &y;
run;

ods output close;
%mend CDR_diff;

%CDR_diff(y=sex);
%CDR_diff(y=ethnic1);

Re: How to export odds ratios, confidence interval and P-value from porc logistic model Macro?

Reeza — Mon, 20 Nov 2017 22:15:03 GMT

That's not what I meant.

I mean, transpose your data before the logistic regression and then use a single PROC LOGISTIC.

Here's a blog post that details the approach.

It's really a better way to go.

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