---
title: "README: miWQS"
date: "`r Sys.Date()`"
geometry: margin=1in
preamble: >
  \usepackage{indentfirst}
  \usepackage{amsmath}
  \usepackage{graphicx}
output: 
   rmarkdown::pdf_document: 
       toc: false
vignette: >
  %\VignetteIndexEntry{README}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
biblography: README.bib
link-citations: true
citation_package: biblatex
csl: multidisciplinary-digital-publishing-institute.csl 
---

<!--- You'll still need to render `README.pdf` regularly, to keep `README.md` up-to-date.-->

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

<!-- The goal of miWQS is to ... ---> 

## Introduction

 We have integrated WQS regression into the MI framework in a flexible $\mathcal{R}$ package called **miWQS** to meet a wide variety of needs (**Figure 1**). The data used in this package consist of a mixture of correlated components that share a common outcome while adjusting for other covariates. The correlated components in the set, $X$, may be complete or interval-censored between zero and low thresholds, or detection limits, that may be different across the components. The common outcome, $y$, may be modeled as binary, continuous, count-based, or rate-based and can be adjusted by the `family` and `offset` arguments of `estimate.wqs()`. Additional covariates, $\boldsymbol{Z}$, may be used in the bootstrap imputation and WQS models. 
 
 If $X$ is interval-censored, the choice of the imputation technique depends on the majority vote of BDL values among the components [1]<!--[@hargartenAccountingUncertaintyDue2020]-->. When most chemicals have 80% of its values BDL, we suggest to use the BDLQ1 approach. When most chemicals have less than 80% of its values BDL, the user should perform Bayesian or bootstrapping multiple imputation. Previous literature suggests to ignore any chemicals that have greater than 80% of its values BDL. The **miWQS** package, though, still allows the user to perform single imputation. Regardless of the technique used, researchers may use the **miWQS** package in order to detect an association between the mixture and the outcome and to identify the important components in that mixture.
 
```{r fig.decide, fig.cap="A decision tree to help researchers in using the miWQS package. The package is flexible and can meet a wide range of needs.", echo = FALSE}
knitr::include_graphics("Decision_Tree_In_Using_miWQS_Package.pdf")
```

## Installation

You can install the released version of miWQS from [CRAN](https://CRAN.R-project.org) with:

```{r eval = FALSE}
install.packages("miWQS")
```

## Example
Please see the vignette [2] <!--[@hargartenRJournal2021]--> for a detailed step-by-step guide in using this package.

## References
<!-- Multidisciplinary Digital Publishing Institute --> 

1. Hargarten, P.M.; Wheeler, D.C. (2020). Accounting for the Uncertainty Due to Chemicals below the Detection Limit in Mixture Analysis. Environmental Research, 186: 109466. https://doi.org/10.1016/j.envres.2020.109466.

2. Hargarten, P.M. & Wheeler, D.C. (2021). miWQS: Multiple Imputation Using Weighted Quantile Sum Regression. The R Journal, 12(2), 226--250. https://doi.org/10.32614/RJ-2021-014