Using Geom_Smooth in R: Customizing Regression Lines

=====================================================

In this article, we will explore how to customize regression lines using the `geom_smooth()` function from the ggplot2 package in R. Specifically, we'll focus on modifying the shape and appearance of smooth line models.

Before diving into the examples, make sure you have installed and loaded the required packages. You can install them via:

```r

install.packages('ggplot2')

```

Then load `ggplot2` using:

```r

library(ggplot2)

```

Understanding geom_smooth()

--------------------------

The `geom_smooth()` function is a powerful tool in ggplot for creating smooth lines on your plots. It allows you to specify various types of smoothing, including linear models (lm), loess functions, and gam models.

In this article, we'll be using the `method` argument within `geom_smooth()`, which specifies how the line should be smoothed. By default, it uses a loess smoother for continuous data.

Specifying Different Colors

--------------------------

One of the most common ways to differentiate between multiple regression lines is by changing their color. You can do this inside or outside the aes function using different arguments within `geom_smooth()`.

Here's an example where we'll use three different linear models (lm) on a dataset called mtcars:

```r

# Load ggplot2 library if not already loaded

library(ggplot2)

# Load mtcars dataset from MASS package. If you don't have it, install and load using:

# install.packages('MASS')

# library(MASS)

load(url('https://raw.githubusercontent.com/hosseinsadeghi/datasets/master/mtcars.csv'))

ggplot(mtcasesses, aes(wt, mpg)) +

# First model: linear

geom_smooth(method = 'lm', formula = 'mpg ~ wt', se = FALSE, color = 'blue') +

# Second model: quadratic (second degree polynomial)

geom_smooth(method = 'lm', formula = 'mpg ~ I(wt^2)', se = FALSE, color = 'red') +

# Third model: cubic (third degree polynomial)

geom_smooth(method = 'lm', formula = 'mpg ~ wt + I(wt^3)', se = TRUE,

aes(color = factor(model)), linetype=1) +

labs(x='Weight', y='Miles per Gallon')

```

In this example:

* We've loaded the mtcars dataset directly from a url since it's not included with R.

* Inside `geom_smooth()`, we use different formulas to create our three models:

* The first model is just linear: `mpg ~ wt`

* The second model adds an interaction term for weight squared: `mpg ~ I(wt^2)`

* The third model includes both the main effect of weight and its cube.

* We've also added a legend using `labs()` to differentiate between our three models.

Customizing Line Appearance

---------------------------

There are several other arguments you can use within `geom_smooth()` to customize how your lines look. Some common ones include:

* linetype: This argument allows you to specify the line style for each model.

* size`: You can change the width of your regression lines using this parameter.

Here's an example where we'll modify some of these aesthetics in our previous code:

```r

ggplot(mtcasesses, aes(wt, mpg)) +

# First model: linear with blue color and solid line style

geom_smooth(method = 'lm', formula = 'mpg ~ wt', se = FALSE,

linetype=1, size=0.5, color='blue') +

# Second model: quadratic (second degree polynomial) in red

# with dashed line style and medium-sized lines.

geom_smooth(method = 'loess', formula = 'mpg ~ wt + I(wt^2)', se = TRUE,

linetype=2, size=1.5, color='red') +

labs(x='Weight (tonne)', y='Miles per Gallon')

```

-------

In conclusion, using `geom_smooth()` in ggplot provides a powerful way to visualize relationships between continuous variables while also allowing for customization of various aspects such as line type and appearance. By understanding how these arguments work together with other functions like aes(), you can create highly informative and visually appealing plots that effectively communicate your findings.

Remember always keep it simple when possible but don't be afraid to add more complexity if needed!

Ggplot shape names