Layer points and a fitted line

Data

Datasets::mpg() returns the 234 cars of ggplot2’s mpg data, bundled with phiz as resources/data/mpg.csv. The example uses two of its columns:

displ is the engine displacement in liters, and hwy is the highway fuel economy in miles per gallon.

Code

use Phiz\Datasets;use function Phiz\{    aes, geom_point, geom_smooth, ggplot, labs,};$mpg = Datasets::mpg();$plot = ggplot($mpg, aes(x: 'displ', y: 'hwy')) |>    geom_point(alpha: 0.5) |>    geom_smooth(method: 'lm') |>    labs(        x: 'Engine displacement (L)',        y: 'Highway miles per gallon',    );$plot->save('layers.html');

use Phiz\Datasets;use Phiz\Plot;$mpg = Datasets::mpg();$plot = Plot::of($mpg)    ->aes(x: 'displ', y: 'hwy')    ->point(alpha: 0.5)    ->smooth(method: 'lm')    ->labs(        x: 'Engine displacement (L)',        y: 'Highway miles per gallon',    );$plot->save('layers.html');

Runnable script: layers.php (PHP 8.5+) or layers.chain.php (PHP 8.3–8.4).

Result

• Inheritance. Both layers use the data and the x and y mappings given to ggplot(). A layer’s own mappings or data override them.
• The band. The band is a 95% confidence interval for the mean response, not a prediction interval for individual cars.
• R parity. phiz computes the line and the band in PHP. They match R’s predict(fit, interval = "confidence").
• Straight line. The points rise again at the largest displacements, which a straight line cannot follow. The default method, loess, fits a local curve (Fitted curves).
• Association. Cars with larger engines tend to have lower highway economy in this data. The line describes that association, not the effect of changing a car’s engine.

In R: ggplot(mpg, aes(displ, hwy)) + geom_point(alpha = 0.5) + geom_smooth(method = "lm").

See also: ggplot() · aes() · geom_point() / ->point() · geom_smooth() / ->smooth() · labs() · save()