Timeline for Parameter Inference when Model is a bad fit to the data.

5 events

when toggle format	what		by	license	comment
Dec 14, 2015 at 4:31	comment	added	erccarls		Now, we fit the model using by maximizing the poisson log-likelihood, and compare models using likelihood ratio tests, where $\Delta\chi^2=-2\Delta\ln(\mathcal{L})$. The problem is that this assumes that the biggest errors are statistical while there are large systematics which are very difficult to calculate. So I am looking for possible ways to characterize the unknown uncertainties without calculating a huge number of models (these are expensive simulations to run).
Dec 14, 2015 at 4:27	comment	added	erccarls		Thanks for the response. I should have been more specific about what I am already doing and the data. The binning is such that the average number of counts per pixel is ~10 or so, depending on which part of the sky we are looking at. In any case, zero pixels are quite rare even with this fine binning.
Dec 11, 2015 at 14:51	history	edited	user32398	CC BY-SA 3.0	added 4 characters in body
Dec 11, 2015 at 2:02	history	edited	user32398	CC BY-SA 3.0	added 279 characters in body
Dec 11, 2015 at 1:55	history	answered	user32398	CC BY-SA 3.0