Log-Likelihoods in light-curve fitting

[darcykenworthy]

According to the manual, the SIGCHI2 term is defined relative to the previous iteration of the fit to prevent numerical uncertainties from subtracting large numbers affecting the fitter. It'd be a useful feature for model verification to be able to evaluate this term relative to the data, rather than relative to the last iteration. Something like:

SIGCHI2_FIXED=2ln(σ (model+data) )- 2ln(σ (data) )