Monday, September 7, 2015
Degenerate Solution Fit for Widmark's 1914 BAC Data
I found the degenerate solution for the diffusion model used in the last posting and tried doing a least squares fit again. I used the trick guessing at where the initial zero, t0, and the final zero, tR, would be as well as "relative likelihood" statistical weights for the sums involved. There was only one parameter, α, to search for since IB was found using the least squares fit. The result was closer to what one would expect from statistical errors in the observations coming within two standard deviations of the data points.
The improvement in the fit came from using the statistical weights which were inversely proportional to the standard deviations of the data points.