On 9/28/10 5:16 AM, Daniele Nicolodi wrote:

Hello,

I copied the code of scipy.optimize.curve_fit into my data analysis code to use it on a not so recent versions of scipy available on my systems. However it does not work when I try to fit a simple function that has only one parameter, for example:

def func(x, p1): return x * p1

The problem is that scipy.optimize.leastsq() returns a list of parameters values for cases where the function has more than one parameter it returns a tuple of parameters values, while in the case of one single parameters it returns a numpy scalar value.

This causes an error in the curve_fit _general_function() and _weighted_general_function() where the target function is called:

function(xdata, *params)

This works only if params is a python list or tuple.

Also the return value suffers from the same problem. If the target function has more than one parameter you get a parameters list, otherwise you get a scalar. It would be more consistent to return always a list, also because the covariance is always returned as a matrix.

Please ignore this if that has been resolved in more recent version of scipy, I have acces to 0.7.2 (last available on debian sid).

This was fixed in 0.8; see http://projects.scipy.org/scipy/ticket/1204 Warren