求助：如何平滑直方图上的正态曲线？geom_smooth是否适用？

我已经手动选择数据范围，用ggplot绘制了叠加两条正态曲线的直方图，但刚接触RStudio，不知道怎么平滑曲线。想问下geom_smooth()是否适用，或者有没有其他解决方法？

当前代码 #

#calculate normal curve and set interval
data_m1 <- data %>% filter(masses_kDa >= 0 & masses_kDa <= 150) #change range here
data_m2 <- data %>% filter(masses_kDa >= 300 & masses_kDa <= 900) #change range here

#calculate mean and standard deviation
mean_m1 <- mean(data_m1$masses_kDa)
sd_m1 <- sd(data_m1$masses_kDa)

mean_m2 <- mean(data_m2$masses_kDa)
sd_m2 <- sd(data_m2$masses_kDa)

library(ggplot2)

# Calculate the bin width and the total number of data points
bin_width <- 50
total_points_m1 <- length(data_m1$masses_kDa)
total_points_m2 <- length(data_m2$masses_kDa)

# Plot the histogram
ggplot(cleaned_data, aes(x=masses_kDa)) +
    geom_histogram(color='transparent', fill='red', alpha = 0.5, bins= 50) + # Adjust the number of bins here
    labs(x=' ', y=' ', title=' ') +
    # Add the normal curves to the histogram
    stat_function(fun = function(x) dnorm(x, mean = mean_m1, sd = sd_m1) * bin_width * total_points_m1, col= '#8B0000') +
    stat_function(fun = function(x) dnorm(x, mean = mean_m2, sd = sd_m2) * bin_width * total_points_m2, col= '#8B0000') +
    # Add the mean lines to the histogram
    geom_vline(aes(xintercept = mean_m1), linetype="dashed", color = 'black') +
    geom_vline(aes(xintercept = mean_m2), linetype="dashed", color = 'black') +
    # Ensure there are no gaps between the graph and axis
    scale_x_continuous(expand = c(0,0)) +
    scale_y_continuous(expand = c(0,0)) +
    coord_cartesian(xlim = c(0, 2000))+ # Adjust the cutoff point here
    theme(panel.background = element_blank(), # Make the background clear
          axis.line = element_line(color = "black"))

解答 #

geom_smooth()并不适用于你的场景——它的作用是对原始观测数据做平滑拟合（比如LOESS、线性回归曲线），而你现在是在绘制预先计算好的理论正态分布曲线，所以用它不对路。

要让你的正态曲线更平滑，有两种简单方法：

方法1：调整stat_function的采样点数

stat_function默认只使用100个点来绘制曲线，点数少就会显得不够平滑。只需给它增加n参数，设置更多采样点即可，比如n=1000：

# 修改后的stat_function部分
stat_function(fun = function(x) dnorm(x, mean = mean_m1, sd = sd_m1) * bin_width * total_points_m1, 
              col= '#8B0000', n = 1000) +
stat_function(fun = function(x) dnorm(x, mean = mean_m2, sd = sd_m2) * bin_width * total_points_m2, 
              col= '#8B0000', n = 1000) +

方法2：提前生成曲线数据框，用geom_line绘制

这种方式更灵活，你可以完全控制曲线的取值范围和点数：

# 先生成两条正态曲线的数据集
curve_m1 <- data.frame(
  x = seq(min(data_m1$masses_kDa), max(data_m1$masses_kDa), length.out = 1000),
  y = dnorm(seq(min(data_m1$masses_kDa), max(data_m1$masses_kDa), length.out = 1000), 
            mean = mean_m1, sd = sd_m1) * bin_width * total_points_m1
)

curve_m2 <- data.frame(
  x = seq(min(data_m2$masses_kDa), max(data_m2$masses_kDa), length.out = 1000),
  y = dnorm(seq(min(data_m2$masses_kDa), max(data_m2$masses_kDa), length.out = 1000), 
            mean = mean_m2, sd = sd_m2) * bin_width * total_points_m2
)

# 绘图时替换stat_function为geom_line
ggplot(cleaned_data, aes(x=masses_kDa)) +
  geom_histogram(color='transparent', fill='red', alpha = 0.5, bins= 50) +
  labs(x=' ', y=' ', title=' ') +
  geom_line(data = curve_m1, aes(x = x, y = y), col= '#8B0000') +
  geom_line(data = curve_m2, aes(x = x, y = y), col= '#8B0000') +
  geom_vline(aes(xintercept = mean_m1), linetype="dashed", color = 'black') +
  geom_vline(aes(xintercept = mean_m2), linetype="dashed", color = 'black') +
  scale_x_continuous(expand = c(0,0)) +
  scale_y_continuous(expand = c(0,0)) +
  coord_cartesian(xlim = c(0, 2000))+
  theme(panel.background = element_blank(),
        axis.line = element_line(color = "black"))

两种方法都能让曲线变得更平滑，方法1更简洁，方法2适合需要对曲线做更多自定义调整的场景。

内容的提问来源于stack exchange，提问作者user124413