This entire two weeks was dedicated to improving the flare energy calculation methods, or more specifically the methods for comparing the different sets of test values for the 3 parameters. Since we had the area method and the correlation method as the ideas, we decided to try a combined method where we sort by corr/area_diff, but it didn't work as the correlation varied a lot less relative to the area difference, making the rankings very similar to that of the area method alone. Since the synthetic emission curves with just the area difference was already pretty good, we finalized our comparison method to sort it by the area difference while removing value sets that had low correlation. Another minor change was that we fixed a bug that had messed with the background emission level. In addition to those changes, I decided to add different weighting factors to different channels of lightcurves. This was because sometimes the top 10 were dominated by synthetic emission curves that were good at longer wavelengths but bad at shorter wavelengths, and I learned that the lightcurves at longer wavelengths are more likely to have greater errors so it's better to put more weight on the shorter wavelengths.
After we finalized our methods, I repeated this procedure with different test values to improve the precision and the accuracy of the parameters for about half of the flares I was assigned to look at. Since it was a lot of waiting while the programs were running, I also tried to comment and understand the code for CME energy Chunming provided me with. This week was the most frustrating week in terms of work since checking how the method worked takes hours to run the programs and that effort not always pays off with an improved results, but I was glad that I was able to finalize our method and start working on the actual calculations.
The slides that I used for the week 8 progress report presentation can be found here.