background_functions.Rmd

---
title: "Background functions"
author: "Brian J. Knaus"
date: "June 25, 2018"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


## Background functions


Some R functions may be used frequently throughout analyses yet to new users some of them may appear non-intuitive.
Here we attempt to identify some of these functions and illustrate their use.


## apply()


One of R's shortcomings is that `for()` loops are relatvely slow.
Historically this was a greater problem and recent versions of R have improved the performance of these loops.
Because of the performance cost associated with `for` loops it is generally recommended to avoid them in your R code, particularly when performance is an issue.
If you have a task that requires iteration and you want performance that is greater than a `for()` loop will provide, you should consider `apply()`.
We'll create a matrix to illustrate its use.


```{r}
tmp <- matrix(rep(1:3, times = 3), ncol = 3)
tmp
```


The function `apply()` will 'apply' a function over the rows or columns of a matrix.
Here we'll use the function `sum()` to illustrate this.


```{r}
apply(tmp, MARGIN = 1, sum)
apply(tmp, MARGIN = 2, sum)
```


The `MARGIN` parameter specifies whether to operate on rows or columns.
I remember how 1 and 2 behave by remembering that when we specify the rows of a matrix with the square brackets (`[]`) we use the first position before the comma.
To specify the columns we use the second position or after the comma.


More information on `apply()` can be found by consulting it's manual page.

```{r, eval=FALSE}
?apply
```

## sweep()


The function `sweep()` is similar to `apply()` in that it iterates over the rows or columns of a matrix.
However, it takes the additional parameters `STATS` and `FUN`.
The parameter `FUN` is a function to be used.
This may be an R function or a custom function you've created.
The parameter `STATS` is a vector of data to be used by the function.
Here we use `apply()` to create a mean for each column and then use `sweep()` to divide the values in the matrix by their column mean.


```{r}
my_means <- apply(tmp, MARGIN = 2, sum) / nrow(tmp)
sweep(tmp, MARGIN = 2, STATS = my_means, FUN = "/")
```

More information on `sweep()` can be found by consulting it's manual page.

```{r, eval=FALSE}
?sweep
```