Here are the results of a test of the FFM method similar to the previous
one (below) but now based on 100-image stacks. The improvements are
appreciable: There is almost no bias (5e-5 or 0.02%) in the determination of albedo (top panel), the scatter is less (SD=0.00034 or 0.11%) and the range of useful lunar phases seems to extend to almost +/-120 degrees. The mean bias is 5 times less than the SD so in effect we have no bias now and are determining albedo to within our science goal. Dotted line shows +/-1% from true albedo value in top panel.
Note that the error bars are unrealistic – they are merely the formal values that the fitting method gives under the assumption that all data are statistically independent. We can get empirical error bars instead by doing some Monte Carlo or bootstrapping on a few of the results.
Test for alfa=1.6 remains to be done. More later!
It does look good, doesn’t it?! Over the weekend I am polishing the results, and producing plots for alfa=1.6. The present fits are un-weighted and this has an effect on the small bias there is (especially for single image fits).
Pete! this stuff is going so well! fantastic. I am looking forward to getting back to ES things next week when I start to write the proposal to the ARC (Australian Research Council). Expect to hear lots from me next week! – Chris