intervals.tripls: Inference for Tri-PLS estimates
In SvenSerneels/tripls_r: Trilinear partial least squares regression

Description Usage Arguments Value Author(s) References Examples

Depending if a data set of new cases is provided, this function provides either confidence intervals for the vector of Tri-PLS regression coefficients, or prediction intervals for the set of new cases. The intervals are computed using error propagation based on a local linearization of order 1.

1	intervals.tripls(res.tripls, alpha = 0.05, Xnew)

`res.tripls`	A "tripls" class regression object containing Tri-PLS regression estimates
`alpha`	The significance level at which to compute the intervals (positive numeric)
`Xnew`	Optional. If missing, intervals for the regression coefficients will be calculated. If provided, it should be a three-dimensional numeric array of dimensions n' x p x q, where the dimenions p and q should match those in res.tripls. In that case, predictions and prediction intervals for the n' new cases will be returned.

Returns a list object containing

`intervals`	A data frame containing the prediction interval's lower limit (llim), the predicted values (either called coefficients or y_predicted, depending on the inputs) and the prediction interval's upper limit (ulim)
`df`	The estimated number of degrees of freedom

Sven Serneels, BASF Corp.

S. Serneels, K. Faber, T. Verdonck, P.J. Van Espen, Case specific prediction intervals for tri-PLS1: The full local linearization. Chemometrics and Intelligent Laboratory Systems, 108 (2011), 93-99.

1
2
3

 data(data_3)
 intervals <- intervals.tripls(tripls(data_3$X3,data_3$y3,1),alpha=.05)
 intervals <- intervals.tripls(tripls(data_3$X3,data_3$y3,1),alpha=.05,data_3$X3)