Conflict between 95% confidence interval for risk difference and p-value

[UcheOkoro]

I noticed that the SAS output showed a conflict between the 95% CI for the risk difference and p-value. The 95% CI is statistically significant, but the p-value is not. How can this be explained and interpreted?

[SAS_Rob · Accepted solution]

In general you can expect the confidence intervals on the difference and the Chi-Square test to correspond with one another when you use the same methodology.  You are using the default Wald Method for the risk difference which would correspond most closely (although not exactly) to the Pearson Chi-Square.  

With respect to Fisher's exact test, there is not a direct correspondence to any of the different confidence limits that are available.  I would expect that if you use CL=EXACT then you will get something fairly close, but again they will not always match up exactly.

tables row*col/riskdiff(cl=exact);

exact riskdiff;

You can refer to the documentation for further details.  SAS Help Center: TABLES Statement

[UcheOkoro]

Thank you so much. It worked.