ensemble_logit_linear: Compute a logit-linear ensemble of predictions
In benkeser/cvma: Cross-validation-based maximal association

Description Usage Arguments Details Value

In general, sl_control$ensemble_fn and y_weight_control$ensemble_fn must accept named inputs, pred and weight, which are, respectively, a matrix and a vector. The matrix has columns corresponding to different learner fits (for sl_control$ensemble_fn) or to the different components of the multivariate outcome (for y_weight_control$ensemble_fn).

1	ensemble_logit_linear(pred, weight, l = 0, u = 1, trim = 0.001)

`pred`	A matrix of predictions
`weight`	A vector of ensemble weights (should have same length as number of columns in `pred`)
`l`	The lower scaling factor.
`u`	The upper scaling factor
`trim`	A value to trim logit computations to avoid numerical instabilities.

In this case, the function computes an ensemble of scaled (by l and u) pred on the logit scale and back-transforms to the original scale. The option trim is used to avoid numeric instabilities.

A vector of ensembled values with length equal to the number of rows of pred.

benkeser/cvma documentation built on May 5, 2019, 1:37 p.m.