I am curious to know whether anybody has run across a randomization/permutation/resampling test for detecting differences between dependent response curves or profiles.

I have hourly observational data (ozone levels) from 5 collection sites for each day of the year. I am wanting to test whether the average monthly profiles are statistically different (in some sense). This problem is analogous to a repeated measures design where the experimental units are repreated over every factor.

I want to use a nonparametric/distribution-free test because of the observational nature of the data. I have run across a permutation test for detecting differences between independent response curves, but all of mine are dependent.

Any input would be greatly appreciated.