rcv.glmnet.Rmd

---
title: Introduction to (a)rcv.gmlnet
output:
    rmarkdown::html_vignette:
        toc_float: true
vignette: >
    %\VignetteIndexEntry{Introduction to (a)rcv.glmnet}
    %\VignetteEngine{knitr::rmarkdown}
    %\VignetteEncoding{UTF-8}
bibliography:
    - bibliography.bib
    - rpackages.bib
---

```{r setup, include = FALSE}
Sys.setenv(LANGUAGE = "en")
library("ameld")

knitr::write_bib("glmnet", "rpackages.bib")
```

**Authors**: `r packageDescription("ameld")[["Author"]] `<br />
**Last modified:** `r file.info("ameld.Rmd")$mtime`<br />
**Compiled**: `r date()`

# Introduction

The `ameld` R package extends `glmnet::cv.glmnet` [@R-glmnet; @glmnet2010].
It supports a repeated cross-validation (`rcv.glmnet`) and
a repeated cross-validation to tune *alpha* and *lambda* simultaneously
(`arcv.glmnet`).

# Dataset

The example data set is taken from `?"glmnet::cv.glmnet"`:

```{r dataset}
set.seed(1010)
n <- 1000
p <- 100
nzc <- trunc(p/10)
x <- matrix(rnorm(n * p), n, p)
beta <- rnorm(nzc)
fx <- x[, seq(nzc)] %*% beta
eps <- rnorm(n) * 5
y <- drop(fx + eps)
```

# Repeated Cross-Validation

`rcv.glmnet` is used to run a repeated cross-validation.
The interface is the same as in `cv.glmnet`.

```{r rcv}
set.seed(1011)
# nrepcv should usually be higher but to keep the runtime of the example low
# we choose 2 here
rcvob <- rcv.glmnet(x, y, nrepcv = 2, nfolds = 3)
plot(rcvob)

plot(rcvob, what = "path")
```

Accessing coefficients or predicting works exactly as with `cv.glmnet`:
```{r rcv-coef}
coef(rcvob)
predict(rcvob, newx = x[1:5, ], s = "lambda.min")
```

# Simultaneously tune *alpha* and *lambda*

`arcv.glmnet` is used to run a repeated cross-validation and tune *alpha* and
*lambda*.

```{r arcv}
set.seed(1011)
# nrepcv should usually be higher but to keep the runtime of the example low
# we choose 2 here
# in similar manner we just evaluate a few alpha values
arcvob <- arcv.glmnet(x, y, alpha = c(0, 0.5, 1), nrepcv = 2, nfolds = 3)
```

We could visualise the effect of different *alpha*s across all *lambda*s in
a plot.

```{r arcv-plot}
plot(arcvob)
```

Alternatively we could just look at the "best" *lambda*.

```{r arcv-plot-lambda.1se}
plot(arcvob, what = "lambda.min")
plot(arcvob, what = "lambda.1se")
```

Next we selecting the best autotuned *alpha*:

```{r arcv-sel-rcv-plot-min.error}
i <- which.min.error(arcvob)
plot(arcvob$models[[i]])
```

# Acknowledgment

This work is part of the [AMPEL](https://ampel.care/en/)
(Analysis and Reporting System for the Improvement of Patient Safety through
Real-Time Integration of Laboratory Findings) project.

This measure is co-funded with tax revenues based on the budget adopted by
the members of the Saxon State Parliament.

# Session Information

```{r sessionInfo}
sessionInfo()
```

# References