missing_value_imputation.Rmd

---
title: "Missing value imputation"
author:
- name: Xiaotao Shen (https://www.shenxt.info/)
date: "Created on 2020-04-01 and updated on `r Sys.Date()`"
output:
  html_document:
    df_print: paged
    toc: no
  pdf_document:
    toc: no
vignette: >
  %\VignetteIndexEntry{missing_value_imputation}
  %\VignettePackage{masscleaner}
  % \VignetteEngine{knitr::rmarkdown}
  % \usepackage[utf8]{inputenc}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE, echo=FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  warning = FALSE,
  message = TRUE,
  out.width = "100%"
)
```

# **Introduction**

We can use `masscleaner` for missing value (MV) imputation.

First, we need to prepare samples for `masscleaner`.

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
library(masscleaner)
library(massdataset)
library(tidyverse)
```

# **Data preparation**

Load the data in [previous step](https://tidymass.github.io/masscleaner/articles/filter_variable_outlier.html).

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
load("peak_tables/POS/object")
```

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
get_mv_number(object)
head(massdataset::get_mv_number(object, by = "sample"))
head(massdataset::get_mv_number(object, by = "variable"))

head(massdataset::get_mv_number(object, by = "sample", show_by = "percentage"))
head(massdataset::get_mv_number(object, by = "variable"), show_by = "percentage")
```

# **Impute missing values**

## zero

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
object_zero = 
  impute_mv(object = object, method = "zero")
get_mv_number(object_zero)
```

## KNN

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
object = 
  impute_mv(object = object, method = "knn")
get_mv_number(object)
```


More methods can be found `?impute_mv()`.

# **Note**

If there are blank samples in dataset, we use different method to impute missing
values.

For Blank samples, just use the zero.

For non-Blank samples, just use the knn or other method

Save data for next analysis.

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
save(object, file = "peak_tables/POS/object")
```

# **Session information**

```{r,eval=TRUE,warning=FALSE, R.options="", message=TRUE, cache=TRUE}
sessionInfo()
```