Greetings,
I have the following problem:
I want to compare a "parameter trajectory", i.e. a series of real
numbers (representing equidistant samples of a time-varying parameter)
produced by some "model", to a reference trajectory, measured from the
real world, in order to get a rating of how good the model that produced
the first trajectory is. Ok, so I use the RMS of the difference between
the two trajectories at each sample.
Then I have another model, producing another trajectory, leading to
another RMS-value. Good. Now I want to clame that the two models are
"equally good" on the basis of the RMS-values being similar, and
require
some sort of significance test to support this claim. But I can't assume
normal distribution in this case since the values are squared (or can
I?) so with my limited knowledge of statistics I'm stuck...
Any help is appreciated!
regards
- Jonas
PS I apologize for posting a question entirely unrelated to R but any
R-related answers are of course welcome!
--
Jonas Beskow, Ph.D. Tel: +46 8 790 8965
Centre for Speech Technology Fax: +46 8 790 7854
KTH beskow at speech.kth.se
SE-10044 Stockholm, Sweden www.speech.kth.se/~beskow
