scikits.statsmodels.regression.WLS

class scikits.statsmodels.regression.WLS(endog, exog, weights=1.0)

A regression model with diagonal but non-identity covariance structure.

The weights are presumed to be (proportional to) the inverse of the variance of the observations. That is, if the variables are to be transformed by 1/sqrt(W) you must supply weights = 1/W. Note that this is different than the behavior for GLS with a diagonal Sigma, where you would just supply W.

Methods

whiten
Returns the input scaled by sqrt(W)
Parameters:

endog : array-like

n length array containing the response variabl

exog : array-like

n x p array of design / exogenous data

weights : array-like, optional

1d array of weights. If you supply 1/W then the variables are pre- multiplied by 1/sqrt(W). If no weights are supplied the default value is 1 and WLS reults are the same as OLS.

Notes

If the weights are a function of the data, then the postestimation statistics such as fvalue and mse_model might not be correct, as the package does not yet support no-constant regression.

Examples

>>> import numpy as np
>>> import scikits.statsmodels as sm
>>> Y = [1,3,4,5,2,3,4]
>>> X = range(1,8)
>>> X = sm.add_constant(X)
>>> wls_model = sm.WLS(Y,X, weights=range(1,8))
>>> results = wls_model.fit()
>>> results.params
array([ 0.0952381 ,  2.91666667])
>>> results.t()
array([ 0.35684428,  2.0652652 ])
<T test: effect=2.9166666666666674, sd=1.4122480109543243, t=2.0652651970780505, p=0.046901390323708769, df_denom=5>
>>> print results.f_test([1,0])
<F test: F=0.12733784321528099, p=0.735774089183, df_denom=5, df_num=1>

Attributes

weights array The stored weights supplied as an argument.
See regression.GLS    

Methods

fit() Full fit of the model.
information(params) Fisher information function of model = - Hessian of logL with respect
initialize()
loglike(params) Returns the value of the gaussian loglikelihood function at params.
newton(params)
predict(exog[, params]) Return linear predicted values from a design matrix.
score(params) Score function of model.
whiten(X) Whitener for WLS model, multiplies each column by sqrt(self.weights)

Previous topic

scikits.statsmodels.regression.GLS.whiten

Next topic

scikits.statsmodels.regression.WLS.fit

This Page