Jump to content

Data Mining Algorithms In R/Packages/gausspred/assess prediction

From Wikibooks, open books for an open world

Description

[edit | edit source]

These functions evaluate predictive probabilities with average minus log probabilities, error rate, and average loss for a defined loss function, or calculate calibration table.

Usage

[edit | edit source]

comp_amlp (probs_pred, responses)

comp_er (probs_pred, responses)

comp_loss (probs_pred, y_true, Mloss)

cal_tab (probs_pred, true_y, ix_y, no_cat=10)

Arguments

[edit | edit source]

probs_pred, a matrix of the predictive probabilities, with rows for cases, columns for groups (different values of response).

Mloss, a matrix defining a loss function, with rows for true values, and columns for predicted values.

responses, 'y_true', 'true_y', a vector of true values of response in test cases.

ix_y, the index of column used to produce calibration table.

no_cat, number of categories in producing calibration table.

Value

[edit | edit source]

comp_amlp, returns average minus log probabilities, comp_er returns error rate.

comp_loss, returns average loss, and expected loss.

cal_tab, returns a calibration data frame.