RC SE calculation
andrjohns opened this issue · 2 comments
With the calculation of the standard errors for ratio of counts data, Spiegelhalter gives the formula:
But I see that you're using:
I know that the 0.5 is to handle 0 counts, but would you have a brief explainer for the rest? Is this a more robust standard error? Thanks!
Hi, I've used Spiegelhalter's work as the basis for a lot, but I actually used the implementation of it in guidance used by the UK healthcare regulator, the CQC: https://www.cqc.org.uk/sites/default/files/Intelligent%20Monitoring%20Statistical%20Methodology%20July%20FINAL%20PDF.pdf .
I've not examined the difference particularly, but the rationale is exactly what you've described, it's to handle zeros. The main point here is converting to z-scores after an approximate normal transformation. It's to try and allow a fairly standard treatment after that as it's all relative once we are on the z-score scale.
Ah that makes sense, thanks for the clarification (and the reference!).