LightResponseCurveFitter_getPriorScale.Rd
return the prior distribution of parameters
LightResponseCurveFitter_getPriorScale(thetaPrior,
medianRelFluxUncertainty, nRec, ctrl)
The beta parameter is quite well defined. Hence use a prior with a standard deviation. The specific results are sometimes a bit sensitive to the uncertainty of the beta prior. This uncertainty is set corresponding to 20 times the median relative flux uncertainty. The prior is weighted n times the observations in the cost. Hence, overall it is using a weight of 1 / 20 of the weight of all observations.
However, its not well defined if PAR does not reach saturation. Need to check before applying this prior
a numeric vector with prior estimates of the parameters