skmisc.loess.loess_model¶
- class skmisc.loess.loess_model(p, family='gaussian', span=0.75, degree=2, normalize=True, parametric=False, drop_square=False)¶
Initialization class for loess fitting parameters
- Parameters:
- p
int
Number of variables
- family
str
One of (‘gaussian’, ‘symmetric’) Determines the assumed distribution of the errors. If ‘gaussian’ the fit is performed with least-squares. If ‘symmetric’ is selected, the fit is performed robustly by redescending M-estimators.
- span
float
Smoothing factor, as a fraction of the number of points to take into account. Should be in the range (0, 1]. Default is 0.75
- degree
int
Overall degree of locally-fitted polynomial. 1 is locally-linear fitting and 2 is locally-quadratic fitting. Degree should be 2 at most. Default is 2.
- normalizebool
Determines whether the independent variables should be normalized. If True, the normalization is performed by setting the 10% trimmed standard deviation to one. If False, no normalization is carried out. This option is only useful for more than one variable. For spatial coordinates predictors or variables with a common scale, it should be set to False. Default is True.
- parametricbool | list-of-bools
of
length
p
Indicates which independent variables should be conditionally-parametric (if there are two or more independent variables). If a sequence is given, the values should be ordered according to the predictor group in x.
- drop_squarebool | list-of-bools
of
length
p
Which squares to drop. When there are two or more independent variables and when a 2nd order polynomial(degree) is used, ‘drop_square’ specifies those numeric predictors whose squares should be dropped from the set of fitting variables. If a sequence is given, the values should be ordered according to the predictor group in x.
- p