Monday, January 02, 2006
st: Using -ci- with weights
I am doing a research project on the incidence of appendicitis over the past ten years. The project is based on data from administrative registries, and I am validating the recorded diagnoses by comparing with information in the medical records. Appendicitis diagnoses originate from a number of hospitals, and I validate 10 diagnoses per hospital. I calculate the positive predictive value for each hospital, with 95% CI, using -ci- with the 'binomial' option, but I have problems calculating the 95% CI for all hospitals combined. The thing is, I would like to weigh each hospital's positive predictive value with the hospital's contribution to the total number of appendicitis cases. In other words, the validity of diagnoses given in large-volume hospitals should matter more than the validity of diagnoses given in hospitals with only few cases. Unfortunately, I am not sure how to do this. Using -ci- with the number of cases per hospital as 'aweight' gives results that look reasonable to me (although the upper limit exceeds 100% on a few occasions), but it seems that I am in effect weighing with each hospital's variance, not the number of cases. I am not really comfortable doing this (perhaps I would feel better if I were a statistician), and I hope that you will give me some ideas to a better approach, if there is one.
Thank you in advance for any assistance, Peter.
* * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/
Links to this post: