glidertools.calibration.robust_linear_fit

glidertools.calibration.robust_linear_fit(gld_var, gld_var_cal, interpolate_limit=3, return_figures=True, **kwargs)

Perform a robust linear regression using a Huber Loss Function to remove outliers. Returns a model object that behaves like a scikit-learn model object with a model.predict method.

Parameters:

  • gld_var (np.array, shape=[n, ]) – glider variable

  • gld_var_cal (np.array, shape=[n, ]) – bottle variable on glider indicies

  • fit_intercept (bool, default=False) – forces 0 intercept if False

  • return_figures (bool, default=True) – create figure with metrics

  • interpolate_limit (int, default=3) – glider data may have missing points. The glider data is thus interpolated to ensure that as many bottle samples as possible have a match-up with the glider.

  • **kwargs (keyword=value pairs) – will be passed to the Huber Loss regression to adjust regression

Returns:

model – A fitted model. Use model.predict(glider_var) to create the calibrated output.

Return type:

sklearn.linear_model