Number of free parameters in an lmfit.Model

[eendebakpt]

I would like to know the number of free parameters in a Model (used to check the input data for fitting has enough data points to perform the fitting). Is there functionality for this already?

If not, would adding a property like:

    @property
    def degrees_of_freedom(self) -> int:
        """ Return number of free parameters in the model """
        fixed_parameters = [ key for key, value in self.param_hints.items() if  value.get('vary', True) is False]
        return len(self.param_names) - len(fixed_parameters)

to the lmfit.Model be a good way to determine the number?

[reneeotten]

See the documentation, it’s already an attribute of the MinimizerResult —> nvarys  https://lmfit.github.io/lmfit-py/fitting.html#lmfit.minimizer.MinimizerResult Renee

[eendebakpt]

@reneeotten The MinimizerResult is only available after fitting right? I want to use the number of degrees of freedom to perform a check on the data before calling Model.fit

[newville]

Neither the number of data points nor the number of free parameters is (or can be) known by a Model until a fit is started.

One might be able to count the number of variable parameters from a Parameters object, which would just be

nvarys = len([p for p in params.values() if p.vary])

That seems simple enough that I'm not sure that needs to be a property of Parameters, but I don't see a lot of harm in adding that.

It definitely does not belong in Model.

[newville]

@eendebakpt See also https://lmfit.github.io/lmfit-py/model.html#lmfit.model.nfree: nfree = ndata - nvarys is the number of degrees of freedom for the fit. Both nvarys and ndata are explicitly stated in every fit report.