Hi Martin I would love to have a friend who was a real mathematician. Does he/she want another friend ? :-)
Using the Sprott algorithm to generate a prior doesn't make any difference to the form of the results: ie. you will still be deriving a posterior probability distribution. Since it is a Bayesian analysis you need a prior of some sort, even it is just a vague / uninformative distribution. Using the Sprott algorithm would represent the statement: I hypothesize the relationship between the missing data and known data to be approximately fractal. As with any Bayesian analysis, the influence of this prior on your results will depend on the strenth of the relationship between the variable of interest and the covariates: - weak relationship => prior will dominate the predictions - strong relationship congruent with prior => prior and covariates evident in results - strong relationship but not conforming to prior => results driven by covariates Michael > Thanks for your suggestion, I didn't though about mixing the two approachs. I > was thinking about implementing them separatly for now. The main reason why > I'm > interrested in the Bayesian approach is to get (for each pixel) the result as > a > probability distribution instead than a single scalar value. We would need to > check if mixing the two approachs would not cause the lost of this probability > distribution. > > Anyway, for anything related to the Bayesian approach, I could not implement > it > alone or answer questions by myself. I will need the help of a friend on mine > who is a real mathematician. We had the plan to implement Bayesian approach > since a few years, but both of us are still missing time up to date... ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ Geotools-devel mailing list Geotools-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geotools-devel
