Let's repeat the curve fit for the second example in the previous page, but instead of guessing "b

Let's get the rising edge data, and make it start at t=0.

and fit the data to the equation:

Last time we guessed a=1, b=-2 and got a good fit to the data. This time let's guess a=1, b=2 and see what happens.

Now, do the fit.

Clearly, the fit is not very good. If we look at the final result, we see that Matlab gets a=-2737 and b=0.000166. Clearly this is wrong and not nearly as good a fit as before.

This illustrates the importance of choosing good initial guesses when doing a non-linear fit.