Computing Power and Sample Size for Quantile Regression

[solfay243]

Hi folks,

I'm looking at the number of widgets someone will obtain regressed on the person's current office level (ordinal) and was wondering if people had code or knowledge of how to best estimate power/sample size for a quantile regression? Google hasn't been much of a help in that department so far. If there are many parameters outside of power, alpha, and number of widget estimates by group then I'd appreciate a quick overview or resources for that as well.

If this is all too much trouble then I've also considered using a log-transformed outcome as well - just makes interpretation and power calculation a tad more complicated.

[Rick_SAS]

Short answer: I don't think SAS provides a procedure for estimating power in quantile regression.

Longer answer: You need to specify a statistical test. For what test are you trying to estimate the power or sample size?

Much longer answer: You can use simulation to estimate the power of any statistical test.