Fit a quadratic polynomial and create a report:

Extract a single property:

Extract several properties at once:

List the available property names:

Fit a linear model and create a report:

Retrieve the best-fitted model:

Get its symbolic expression:

Get a pure function ready to evaluate:

Or a function that takes a row or list input:

Fit a linear model on a tabular dataset with two predictors:

Inspect the model name, hyperparameters, variables and input/output sizes:

View the data type schema:

Inspect the inferred task and the column-to-variable mapping:

Fit a quadratic polynomial and create a report:

Inspect the fitted parameter values:

Or as a name-value association:

Or as a list of rules:

Inspect the number of parameters:

Fit a polynomial model on a noisy sinusoid:

Inspect the training loss and the loss function used:

Inspect summary statistics of the fit:

View the fit residuals:

Fit a linear model and create a report:

Residual histogram with a fitted normal overlay:

Residuals against fitted values:

Residuals against the predictor (1D inputs only):

Compare the residual distribution to a normal distribution:

Predicted versus actual response:

Compare three candidate models on the same data:

View the cross-validation chart of per-fold losses:

Inspect per-candidate cross-validation results:

Inspect the validation set strategy used:

Tune a model hyperparameter automatically:

Inspect the searched and selected hyperparameters:

View the cross-validation chart of candidate degrees:

Fit a parametric polynomial and create a report:

View standard goodness-of-fit measures:

View the analysis-of-variance table:

View parameter estimates with standard errors, t-statistics, p-values and confidence intervals:

View the parameter correlation matrix:

View confidence bands for the mean prediction and for individual observations:

View the fitted function with its mean uncertainty:

Inference is also available for nonlinear "Formula" models:

Inspect parameter estimates of the nonlinear fit:

Fit a cubic polynomial and create a report:

View the basis used in the fit:

View the design matrix:

View Cook's distance influence measures:

View the Durbin–Watson statistic for residual autocorrelation:

View variance inflation factors for collinearity diagnostics:

View the eigenstructure of the design matrix:

By default, multiple-candidate fits use cross-validation to select the best:

Use ValidationSetNone to disable cross-validation; selection then uses training loss, which favors more complex models:

Compare candidate calibration polynomials by cross-validation:

Inspect parameter estimates with confidence intervals for the selected calibration:

Check that the residuals are consistent with the noise model: