P values and ttest

[Michael Andersonnnn]

please forgive my naivety here, but I am having issues interpreting the p-values when doing ttests in STATA, I am finding that I have overlapping 95% confidence intervals but my p value is less than 0.05

The code i am using is

ttest avgdiagtotal, by (secondaryanalysis)

The output I get is attached.



You can see that Pr(|T| > |t|) = 0.0192
I would expect this p value to be more than 0.05, as the confidence intervals overlap??

[Clyde Schechter]

Your intuitions are leading you astray here. Overlapping confidence intervals do not necessarily imply a statistically significant difference. The reason is that the widths of the confidence intervals are based on a different standard error calculation than method for calculating the standard error of the difference. If you would like more details and a fully worked example you can see https://www.google.com/url?sa=t&rct=...Wr3eixyM8TOu9n. (PDF download).

[Carlo Lazzaro]

Michael:
as an aside to Clyde's towering explanation, you may want to read: https://www.wiley.com/en-us/Statisti...-9780470144480 pages 38-40.

[Maarten Buis]

Yet another text discussing this: http://www.stat.columbia.edu/~gelman...ed/signif4.pdf