Statistical Tests for Residual AnalysisΒΆ

This is a collection of statistical tests that can be used independent from any models. They are used in :mod: statsmodels to perform tests of the underlying assumptions of a statistical model, e.g. durbin_watson for autocorrelation of residuals, jarque_bera for normal distribution of residuals. Further tests on the properties of the residuals can be obtained from scipy.stats.

durbin_watson(resids) Calculates the Durbin-Waston statistic
jarque_bera(resids) Calculate residual skewness, kurtosis, and do the JB test for normality
omni_normtest(resids[, axis]) Omnibus test for normality
conditionnum(exog) Returns the condition number of an exogenous design array.

Previous topic

scikits.statsmodels.discretemod.DiscreteResults.conf_int

Next topic

scikits.statsmodels.stattools.durbin_watson

This Page