predict_lowess

Adapted from <code>rms:::calibrate.default</code>
Uses <code>approx</code> with <code>rule</code> = 2 so that x out of range of initial lowess fit returns min or max (see ?approx)

internal

Fit calibrations curves for clinical prediction models and calculate several associated
metrics (Eavg, E50, E90, Emax). Ideally predicted probabilities from a prediction model
should align with observed probabilities. Calibration curves relate predicted probabilities
(or a transformation thereof) to observed outcomes via a flexible non-linear smoothing function.
'pmcalibration' allows users to choose between several smoothers (regression splines, generalized
additive models/GAMs, lowess, loess). Both binary and time-to-event outcomes are supported.
See Van Calster et al. (2016) <doi:10.1016/j.jclinepi.2015.12.005>;
Austin and Steyerberg (2019) <doi:10.1002/sim.8281>;
Austin et al. (2020) <doi:10.1002/sim.8570>.

Stephen Rhodes

pmcalibration

Calibration Curves for Clinical Prediction Models

predict_lowess function

<dl><dt>fit</dt>
<dd>list produced by <code>lowess</code></dd>
<dt>x</dt>
<dd>values to produce predictions for</dd></dl>

Arguments

Get predictions from loewss fit — predict_lowess

<dl>

<dt>fit</dt>
<dd>list produced by <code>lowess</code></dd>


<dt>x</dt>
<dd>values to produce predictions for</dd>

</dl>

predict_lowess: Get predictions from `loewss` fit

Description

Usage

Value

Arguments