I am planning to repeat a published analysis using a larger sample. I wish to estimate the impact of my better sample size on the precision of my estimate of a hazard ratio.
I can do this for a rate ratio, on the assumption that the point estimate will be similar and that the number of events will scale with the sample size
Thanks,
Tom
I can do this for a rate ratio, on the assumption that the point estimate will be similar and that the number of events will scale with the sample size
However, all of these studies report hazard ratios rather than rate ratios, which is what I plan to report too. Is there straight forward way to calculate how precision of a hazard ratio would scale with sample size. Is there a STATA command that will estimate this? I can't see that ciwidth or ciwidth_proportions_mcwill do this, though I may be wrong?
SE of log(rate ratio) = √ ( 1 / events in exposed + 1 / events in unexposed)
EF = exp[1.96 x SE of log(rate ratio)]
95% confidence interval = [rate ratio / EF] to [rate ratio * EF]
Thanks,
Tom
