topic P-value and 95% CI disagreement in Statistical Procedures

P-value and 95% CI disagreement

GS2 — Fri, 03 Jan 2020 20:46:55 GMT

Using SAS 9.4

I ran the following code

PROC LOGISTIC DATA =have;
CLASS tertile (REF = '0') ;
MODEL infection (event= 'Yes')= tertile / stb;
RUN;

I have binary infection variable and tertile is categorical with 3 levels.

My question is when running this code it returned a p-value of 0.04 and an OR 7.1 and a 95% CI of (0.874, 58.689). Can this be correct to have an CI that includes 1 and a significant p-value?

Re: P-value and 95% CI disagreement

Rick_SAS — Fri, 03 Jan 2020 21:21:28 GMT

Did you notice that the Type 3 Test shows that the main effect (Tertile) is not significant for predicting Infection?

Re: P-value and 95% CI disagreement

PGStats — Fri, 03 Jan 2020 21:22:16 GMT

Try adding option CLPARM=BOTH to the model statement. From proc logistic documentation:

There are two methods of computing confidence intervals for the regression parameters. One is based on the profile-likelihood function, and the other is based on the asymptotic normality of the parameter estimators. The latter is not as time-consuming as the former, because it does not involve an iterative scheme; however, it is not thought to be as accurate as the former, especially with small sample size.

Re: P-value and 95% CI disagreement

GS2 — Fri, 03 Jan 2020 21:25:40 GMT

I saw that which was why I was wondering that the individual tertile is showing significance. Should I have used a wald test in this situation? Thank you

Re: P-value and 95% CI disagreement

StatDave — Mon, 06 Jan 2020 16:04:45 GMT

Different things are being compared. See this note.