T-Test - how is p <0.05 but 95% confidence intervals overlapping for mean

[Gaurav Mohan]

Here we see that the alternate hypothesis that difference != 0 has p=0.03; How is this possible when the 95% CI are overlapping? Please explain

In this test I was trying to find if the mean of ast (blood test value with a numerical value) is significantly different for those who went to the icu (Intensive Care Unit) or not.

I used the code:
test alt, by(icu)

[Leonardo Guizzetti]

The test statistic is related to the confidence interval around the difference. Said differently, it's possible for the confidence interval for two statistics to overlap yet still have a statistically significant difference.

[Felix Bittmann]

https://link.springer.com/article/10...423-015-0947-8
https://journals.sagepub.com/doi/pdf...10581001900316
https://www.graphpad.com/guides/pris..._significa.htm