LinearRegressionResult

LinearRegressionResult(**fields)

Fitted ordinary least squares model, mirroring the C# LinearRegression class.

Carries no C++ state: predict reruns the fit against the shared toolbox runner each call, so an instance serializes with :mod:pickle.

covariance is the coefficient covariance matrix, i.e. numpy.sqrt(numpy.diag( result.covariance)) equals result.standard_errors. The underlying C# LinearRegression.Covariance is the unscaled cross-product term ((X'X)^-1), scaled here by sigma**2 to match standard_errors.

Methods

Name Description
predict Predict from the fitted model.

predict

LinearRegressionResult.predict(newdata, interval=False, level=0.9)

Predict from the fitted model.

Mirrors the C# LinearRegression.Predict/PredictionIntervals methods.

Parameters

Name Type Description Default
newdata array_like A 2D array of predictors with the same number of columns the model was fitted with (an intercept column, if any, is added internally), or – for a single-predictor model only – a 1D array of observations (matching R’s predict.corehydro_lm(), where a bare vector is unambiguous when there is one predictor). For more than one predictor a 1D array is rejected rather than guessed at, since it is ambiguous whether it means one row or one column. required
interval bool If True, also return the level prediction interval (a Student-t interval around the mean response, mirroring PredictionIntervals). Default False. False
level float Prediction interval level, between 0 and 1. Default 0.90, matching PredictionIntervals’s own alpha=0.1 default. 0.9

Returns

Name Type Description
numpy.ndarray With interval=False, one predicted value per row of newdata. With interval=True, an (n, 3) array with columns lower, upper, mean.