Calcite | Level 5

## CIs on Regression Coefficients from SAS Proc Logistic

All-

I ran a simple logistic regression in SAS to obtain the CIs on the regression coefficients. The CIs do not appear to be correct. Can anyone let me know what they think is going on. If you calculate the CIs by hand, these do not match what is provided in the SAS output.

 Analysis of Maximum Likelihood Estimates Parameter DF Estimate StandardError WaldChi-Square Pr > ChiSq Intercept 1 7.6653 22.9839 0.1112 0.7388 x 1 -1.1236 3.3681 0.1113 0.7387

 Parameter Estimates and Wald ConfidenceIntervals Parameter Estimate 95% Confidence Limits Intercept 7.6653 -37.3823 52.7128 x -1.1236 -7.7250 5.4779

By hand, I get the 95% (2-sided) CI for the intercept: (-52.71234 ,  37.38174 )

SAS Code:

proc import datafile="C:\Users\BodnarJ\Desktop\functionalRequirement4_9_X\dataSim_FR_4_9_X.csv" dbms=csv out=work.anaDS replace;
delimiter="|";
getnames=yes;
guessingrows=7000;
run;

proc logistic data = anaDS ;
model yBinom1 = x / cl;
output out=propensity_scores pred=prob_flag;
run;

1 ACCEPTED SOLUTION

Accepted Solutions
Opal | Level 21

## Re: CIs on Regression Coefficients from SAS Proc Logistic

You must be modelling for the yBinom1=0 events. Try the syntax

``model yBinom1(event='1') = x / cl;``
PG
3 REPLIES 3
Diamond | Level 26

## Re: CIs on Regression Coefficients from SAS Proc Logistic

Typically, the actual estimated coefficient for a logistic regression is approximately half-way between the upper confidence interval and the lower confidence interval.

Therefore, I conclude that your hand calculations are wrong.

--
Paige Miller
Opal | Level 21

## Re: CIs on Regression Coefficients from SAS Proc Logistic

You must be modelling for the yBinom1=0 events. Try the syntax

``model yBinom1(event='1') = x / cl;``
PG
Obsidian | Level 7

## Re: CIs on Regression Coefficients from SAS Proc Logistic

The above solution is absolutely correct!

Discussion stats
• 3 replies
• 1409 views
• 2 likes
• 4 in conversation