Logistic Fit

[Tim Wessman]

Well, this logistic fit was pulled from the old hp math library, but frankly it was never any good due to being so sensitive to even minor changes in numbers. When I was reimplementing it for the 39gII, I really wanted to switch it for a much better and more robust algorithm. However, after many many fruitless days of searching (over many months) I was never able to find a good fit that behaved predictably. The biggest challenge with this type of fit is finding good initial estimates. A human can easily identify what a reasonable estimate is, and whether it should be an increasing/decreasing version, but finding an algo that matched what is desired proved ridiculous difficult.

If anyone has any recommendations or suggestions, I am totally welcome to them.