README.Rmd

---
output: github_document
---

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

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  dpi = 400,
  out.width = "100%"
)
```

# fwlplot

<!-- badges: start -->
<!-- badges: end -->

This is a super simple package to help make scatter plots of two variables after residualizing by covariates. This package uses `fixest` so things are super fast. This is meant to (as much as possible) be a drop in replacement for `fixest::feols`. You should be able to replace `feols` with `fwl_plot` and get a plot.

## Installation

You can install the development version of fwlplot like so:

``` r
devtools::install_github("kylebutts/fwlplot")
```

## Example

Here's a simple example with fixed effects removed by `fixest`. 
```{r}
library(fwlplot)
library(fixest)

flights <- data.table::fread("https://raw.githubusercontent.com/Rdatatable/data.table/master/vignettes/flights14.csv")
flights$long_distance = (flights$distance > 2000)
# Sample 10000 rows
sample = flights[sample(nrow(flights), 10000), ]
```

```{r}
#| fig.width:  8
#| fig.height: 4
# Without covariates = scatterplot
fwl_plot(dep_delay ~ air_time, data = sample)
```

```{r}
#| fig.width:  8
#| fig.height: 4
# With covaraites = FWL'd scatterplot
fwl_plot(
  dep_delay ~ air_time | origin + dest, 
  data = sample, vcov = "hc1"
)
```

### Plot random sample

If you have a large dataset, we can plot a sample of points with the `n_sample` argument. This determines the number of points *per plot* (see multiple estimation below).

```{r}
#| fig.width:  8
#| fig.height: 4
fwl_plot(
  dep_delay ~ air_time | origin + dest, 
  # Full dataset for estimation, 1000 obs. for plotting
  data = flights, n_sample = 1000
)
```

### Full `feols` compatability

This is meant to be a 1:1 drop-in replacement with fixest, so everything should work by just replacing `feols` with 


```{r}
#| fig.width:  8
#| fig.height: 4
feols(
  dep_delay ~ air_time | origin + dest, 
  data = sample, subset = ~long_distance, cluster = ~origin
)
```
```{r}
#| fig.width:  8
#| fig.height: 4
fwl_plot(
  dep_delay ~ air_time | origin + dest, 
  data = sample, subset = ~long_distance, cluster = ~origin
)
```


### Multiple estimation 

```{r}
#| fig.width:  8
#| fig.height: 4
# Multiple y variables
fwl_plot(
  c(dep_delay, arr_delay) ~ air_time | origin + dest, 
  data = sample
)
```

```{r}
#| fig.width:  8
#| fig.height: 6
# `split` sample
fwl_plot(
  c(dep_delay, arr_delay) ~ air_time | origin + dest, 
  data = sample, split = ~long_distance, n_sample = 1000
)
```

```{r}
#| fig.width:  8
#| fig.height: 10
# `fsplit` = `split` sample and Full sample
fwl_plot(
  c(dep_delay, arr_delay) ~ air_time | origin + dest, 
  data = sample, fsplit = ~long_distance, n_sample = 1000
)
```

### ggplot2

```{r}
#| fig.width:  8
#| fig.height: 10
library(ggplot2)
theme_set(theme_grey(base_size = 16))
fwl_plot(
  c(dep_delay, arr_delay) ~ air_time | origin + dest, 
  data = sample, fsplit = ~long_distance, 
  n_sample = 1000, ggplot = TRUE
)
```