Confidence intervals

[Christos Chalitsios]

Hello! I am using a Cox proportional hazard model to examine the association between LDL-c and a disease using robust SE (cluster(practice). I noticed that the 95% CI are not symmetric. Why does this happen?

Many thanks!

[Maarten Buis]

This is discussed here: https://www.stata.com/support/faqs/s...cs/delta-rule/ . In short, that non-symmetry of the confidence interval is desirable: the sampling distribution of the logarithm of the hazard ratio is approximately normally distributed, and its confidence interval is symmetric. However, the sampling distribution of the hazard ratio is thus not symmetric, and the confidence interval reflects that.