If I have two risk ratios (calculated from Poisson regression) and 95% confidence intervals for each ratio, can I make the claim that the risk ratios are significantly different if there is no overlap between the confidence intervals?

Moreover, if I can make that claim, is there a citation that proves that that is a legitimate claim to make?

  • $\begingroup$ I don't get what you mean by "two risk ratios". In Poisson regression, we model the expected value of counts on the basis of features. $\endgroup$
    – utobi
    Feb 14 at 9:26
  • $\begingroup$ Comparing two confidence intervals based on the visual overlap is never legitimate in any circumstance. $\endgroup$
    – wzbillings
    Feb 14 at 14:54