power for fracreg, betareg, zoib

[Rich Goldstein]

any help on computing power/sample size (and/or detectable effect size) for an outcome variable that is a proportion would be appreciated, including cites to articles

context: for a proposed study examining the effect of burns on range of motion (ROM) at body joints, we are not using actual ROMs (since the possible ranges vary widely by joint (and sometimes also by type of motion)) and so we are making these commensurate by using the proportion of the maximum ROM as our outcome variable; I am told that both 0 and 1 are possible but whether they occur in the data set to be used is not yet clear

forgot to mention that -zoib- is user written (Maarten Buis) and is available from SSC

[Maarten Buis]

I have never considered the power of these methods, but if I were to do so I would probably do it through simulation: https://www.stata.com/support/faqs/s...by-simulation/