Nonlinear Regression Classes

Suggestions about class inheritance to be followed during the course of project -

NonLinearModel - Base Class for all non linear models. Derived from Class Model
Derived Classes-

• NonlinearLS - Nonlinear Least Squares class similar to OLS and WLS. Use Levenberg-Marquardt Algorithm for curve fitting. Currently using scipy.optimize.leastsq() for this purpose.
• NonlinearRobust - Class for calculating nonlinear robust estimators. IRLS algorithm or its variant used for calculation.

NonLinearRegressionResults -Base Class for holding the estimates and predictions. Derived from Regression Results
Derived Classes-

• NonlinearLSResults - For holding results from fitting NonlinearLS
• NonlinearRobustResults - For holding results from NonlinearRobust. Furthur classes may be derived for different estimators.

The user will have to provide an expression in the expr method of a class derived from the respective NonLinear classes for fitting the data.
Optional expressions will be provided.