The values of "a" known as beta in the functions help are the initial values that the model will use as parameters in your 'modelfun'.
The values are 100% dependent on your model - you can start with zeros, ones or values which are "likely" to be near the end result - it can effect the solution so think about it carefully in advance.
