README.Rmd

---
title: "Kenward-Roger, Satterthwaite and Parametric Bootstrap Based Methods for Tests in Mixed Models with the `pbkrtest` package"
output: 
  github_document:
    toc: true
    toc_depth: 2
---

<!-- README.md is generated from README.Rmd. Please edit only README.Rmd! -->

```{r, echo = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  message = FALSE
)
```

## What does `pbkrtest` do for you?

Hypothesis test of fixed effects in mixed models (also called random effects models, hierarchical models etc) is most commonly based on large sample asymptotics: When the amount of information becomes large, a test can be based an a $\chi^2$-approximation. In small sample cases, this approximation can be very unreliable. The `pbkrtest` provides alternatives to this approximation. To be specific: For linear mixed models (as implemented in the `lme4`
package), `pbkrtest` implements the following tests for fixed effects:

1. a parametric bootstrap test, 
1. a Kenward-Roger-type F-test and
1. a Satterthwaite-type F-test. 

Moreover, for generalized linear mixed models (as implemented in `lme4`) and for
generalized linear models, `pbkrtest` also implements a parametric
bootstrap test

## Documentation

The facilities of the package are documented in the paper by [Halekoh and Højsgaard 2014)]
(https://www.jstatsoft.org/htaccess.php?volume=059&type=i&issue=09&filename=paper)
Please see `citation("pbkrtest")` for information about citing the
paper and the package. If you use the package in your work, please do
cite this paper.  Please notice: There are other packages that use
`pbkrtest` under the hood. If you use one of those packages, please do
also cite our paper.

We also refer to the [Webpage for the package](https://people.math.aau.dk/~sorenh/software/pbkrtest/index.html) 


<!-- badges: 
[![R build status](https://github.com/hojsgaard/pbkrtest/workflows/R-CMD-check/badge.svg)](https://github.com/hojsgaard/pbkrtest/actions) 
[![codecov.io](https://codecov.io/gh/hojsgaard/dlmextra/branch/master/graphs/badge.svg)](https://codecov.io/gh/hojsgaard/dlmextra?branch=master)
badges: end -->


## Online documentation

See <https://hojsgaard.github.io/pbkrtest/>.


## Installation

`pbkrtest` is available on CRAN and development versions can also be found on Github:

```
## Install from CRAN:
install.packages('pbkrtest')
## Install from Github: Use the remotes package:
remotes::install_github("hojsgaard/pbkrtest", build_vignettes = TRUE)
```

## Development site

See <https://github.com/hojsgaard/pbkrtest>.

## Brief introduction


```{r, message=FALSE}
library(pbkrtest)
library(ggplot2)

## Sugar beets: Does suger content depend on harvest time?

beets |> ggplot(aes(x=sow, y=sugpct, group=harvest)) +
    geom_jitter(aes(color=harvest), width=0)

fm0 <- lmer(sugpct ~ block + sow + harvest + (1|block:harvest), data=beets)
fm1 <- update(fm0, .~. -harvest)

## Is there an effect of harvest time?
an <- anova(fm0, fm1)
pb <- PBmodcomp(fm0, fm1)
kr <- KRmodcomp(fm0, fm1)
sa <- SATmodcomp(fm0, fm1)

tidy(an)
tidy(pb)
tidy(kr)
tidy(sa)
``` 

Please find more examples in the other vignettes available at <https://hojsgaard.github.io/pbkrtest/>.