Test with mock data
I completed a test of the MLE calculation with mock data built from the cube.
I integrated km/s about the velocity center of the cube, used a DGR of cm mag, and an intercept of mag to create the mock data.
The MLE analysis calculated the same parameter values used to create the data as the best estimates. This analysis included masking the data in the iterative manner described in section 3 of the current draft.
So either some assumption in the analysis for the given data is incorrect, or the negative intercepts are real. I’ll have to think more on this.
See line 127 through 165 of this version of the
Testing with different masking techniques
Instead of performing the iterative masking technique, I ran the MLE analysis on the Lee+12 IRIS data. I masked all pixels for mag. The MLE analysis still included the binning and rescaling of uncertainties in the pixels, only the masking step was excluded.
The uncertainty in the likelihood spaces are more of what I would expect from each parameter. This is likely because we are not excluding pixels which do not fit the model very well from the residual masking technique.
Somehow the pixels with a background are being excluded during the masking process, leading to a poor estimate of the true background.
More testing to come…