如何将 log2 转换的可视化合并到函数中？

Question

在这个thread 的帮助下，我拥有了这个norm_f() 函数。该函数生成两个图来评估正态分布：

> norm_f(iris$Sepal.Length)

函数是这样写的：

norm_f <- function(z, hist. = list(col = "lightgray"),
                   lines. = list(lwd = 2), curve. = list(col = "red", lwd = 2),
                   qqnorm. = list(col = "black", pch = 1, cex = 1.2), qqline. = list(lwd = 2)) {
  zname <- deparse(substitute(z))
  zden <- density(z)
  if (!"ylim" %in% names(hist.)) {
    ymax <- max(zden$y, dnorm(c(seq(min(z), max(z), len = 21)), mean = mean(z), sd = sd(z)))
    hist.$ylim <- c(0, ymax)
  }
  par(mfrow=c(1,2), mar=c(5,5,2,2))
  if (!"main" %in% names(hist.)) hist.$main <- paste("Histogram of", zname)
  if (!"xlab" %in% names(hist.)) hist.$xlab <- zname
  do.call(hist, c(list(z,  prob = TRUE), hist.))
  do.call(lines, c(list(zden), lines.))
  do.call(curve, c(list(substitute(dnorm(x, mean=mean(z), sd=sd(z))),
                        add = TRUE), curve.))
  do.call(qqnorm, c(list(z), qqnorm.))
  do.call(qqline, c(list(z), qqline.))
  
}

我想更新函数，使其生成 2x2 图：两个没有 log2 变换（如上）和两个带有 log2 变换协变量的图（即，对应于 norm_f(log2(iris$Sepal.Length))）

预期输出：

Answer 1

我修改了原始函数norm_f() 以使用ggplot2 和patchwork。现在该函数接受三个参数：df 是你的 data.frame，var 是一个代表你的目标变量的字符（例如，Sepal.Length）和thefun，它调用你想要的不带引号的缩放函数。这些图是通过 ggplot2 渲染的，它们的布局是通过 patchwork 组织的。

请查看下面的代码以检查它是否符合您的需求。我运行了两个示例，一个使用请求的log2() 函数，第二个使用exp()。

library(ggplot2)
library(patchwork)

data(iris)

norm_f <- function( df, var, thefun, ... ) {
  
  # this gets the name of the function invoked
  fun_name = as.character(match.call()[4]) 
  # this matches the function in `thefun` so that you can use it
  thefun = match.fun(thefun)
  
  # compute the log2
  va2_log2 <- thefun(df[ , var ])
  
  # specify we want the density for the histogram
  y <- "..density.."
  # Histogram of "var"
  p_hist <- ggplot( df, aes_string(x = var) ) +
    geom_histogram(aes_string( y = y ) ) + 
    geom_density() +
    stat_function(fun = dnorm, 
                  args = list(mean = mean(df[ , var ], na.rm = TRUE), 
                              sd = sd(df[ , var ], na.rm = TRUE)), 
                  colour = 'blue') + 
    ggtitle(paste0("Histogram of ", var) )
  
  # QQ-plot of "var"
  p_qq <- ggplot( df, aes_string(sample = var) ) +
    stat_qq() + stat_qq_line() +
    ggtitle(paste0("QQ-Plot of ", var) )
  
  # Histogram of log2("var")
  p_hist_log <- ggplot( df, aes_string(x = va2_log2) ) +
    geom_histogram(aes_string( y = y ) ) + 
    geom_density() +
    stat_function(fun = dnorm, 
                  args = list(mean = mean(va2_log2, na.rm = TRUE), 
                              sd = sd(va2_log2, na.rm = TRUE)), 
                  colour = 'blue') +
    ggtitle( paste0("Histogram of ", fun_name, "(", var, ")" ) )
  
  # QQ-plot of log2("var")
  p_qq_log <- ggplot( df, aes_string(sample = va2_log2) ) +
    stat_qq() + stat_qq_line() +
    ggtitle( paste0("QQ-Plot of ", fun_name, "(", var, ")" ) )
  
  print((p_hist + p_qq) / (p_hist_log + p_qq_log))
  
}

norm_f(df = iris, var = "Sepal.Length", thefun = log2 )
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

norm_f(df = iris, var = "Sepal.Length", thefun = exp )
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
#> `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

^{由reprex package (v1.0.0) 于 2021-03-01 创建}