In Katsevich-Lab/katlabutils: Utility functions broadly useful for Katsevich Lab research

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
set.seed(1)

We often make QQ plots in the Katsevich Lab, and we've found the qqplotr package to be somewhat deficient for this purpose. In particular, this package breaks when multiple QQ plots are displayed on the same axes, or when axis transformations are used. The katlabutils package therefore contains functions to help us make QQ plots with the aforementioned functionalities.

library(katlabutils)
library(ggplot2)
library(dplyr)
library(tibble)

Let's start with a basic example of a vanilla QQ plot for uniform random variables.

data <- tibble(pvalue = runif(100))
data |>
  ggplot(aes(y = pvalue)) + # variable to be plotted in y aesthetic
  stat_qq_points() +        # add the points to the QQ plot
  stat_qq_band() +          # add the band to the QQ plot
  geom_abline() +           # add the 45 degree line
  labs(
    x = "Expected quantile",
    y = "Observed quantile"
  ) +
  theme_bw()

If we want to look at the tail of the distribution, it's useful to log-transform both axes. We can do this with the help of the revlog_trans() function:

data |>
  ggplot(aes(y = pvalue)) +
  stat_qq_points() +
  stat_qq_band() +
  geom_abline() +
  labs(
    x = "Expected quantile",
    y = "Observed quantile"
  ) +
  scale_x_continuous(trans = revlog_trans(base = 10)) +
  scale_y_continuous(trans = revlog_trans(base = 10)) +
  theme_bw()

We can create QQ plots for arbitrary distributions from the stats package. We just need to specify the distribution to the distribution argument to both stat_qq_points and stat_qq_band. If the stats package has density function dxxxx, then we pass distribution = "xxxx" to stat_qq_points and stat_qq_band. Below is an example of a normal QQ plot:

data <- tibble(zvalue = rnorm(100))
data |>
  ggplot(aes(y = zvalue)) +
  stat_qq_points(distribution = "norm") +
  stat_qq_band(distribution = "norm") +
  geom_abline() +
  labs(
    x = "Expected quantile",
    y = "Observed quantile"
  ) +
  theme_bw()

The stat_qq_points and stat_qq_band layers can be composed arbitrarily with ggplot2's functionalities. For example, we can display multiple QQ plots on the same axes using color:

data <- tibble(
  pvalue = c(runif(100), sqrt(runif(100))),
  method = c(rep("A", 100), rep("B", 100))
)
data |>
  ggplot(aes(y = pvalue, colour = method)) +
  stat_qq_points() +
  stat_qq_band() +
  geom_abline() +
  labs(
    x = "Expected quantile",
    y = "Observed quantile"
  ) +
  scale_x_continuous(trans = revlog_trans(base = 10)) +
  scale_y_continuous(trans = revlog_trans(base = 10)) +
  theme_bw()

We can also using ggplot2's faceting functionality:

data <- rbind(
  tibble(
    pvalue = c(runif(100), sqrt(runif(100))),
    method = c(rep("A", 100), rep("B", 100)),
    dataset = c("Dataset 1")
  ),
  tibble(
    pvalue = c(runif(100), runif(100)),
    method = c(rep("A", 100), rep("B", 100)),
    dataset = c("Dataset 2")
  )
)
data |>
  ggplot(aes(y = pvalue, colour = method)) +
  stat_qq_points() +
  stat_qq_band() +
  geom_abline() +
  scale_x_continuous(trans = revlog_trans(base = 10)) +
  scale_y_continuous(trans = revlog_trans(base = 10)) +
  facet_wrap(~dataset) +
  labs(
    x = "Expected quantile",
    y = "Observed quantile"
  ) +
  theme_bw()

Katsevich-Lab/katlabutils documentation built on Aug. 7, 2024, 4:55 p.m.