计算和绘制 r 中核密度分布的差异

Question

I'm using R and I want to calculate the differences between two Kernel density distributions at each point on the x axis and plot that difference but am having some trouble. Is there a certain function or way that I can do this? For context, I'm using blood pressure data and I want to calculate the differences at each point in the blood pressures between men and women.

My code for the distributions (not the differences) looks something like this (SBP=systolic blood pressure):

km <- density(data$SBP[data$GENDER==0], bw="nrd0", adjust = 1, kernel = c("gaussian"), window = kernel, n=512, cut=3, give.Rkern = FALSE, na.rm=FALSE)
kf <- density(data$SBP[data$GENDER==1], bw="nrd0", adjust = 1, kernel = c("gaussian"), window = kernel, n=512, cut=3, give.Rkern = FALSE, na.rm=FALSE)

plot(km, xlab="SBP", main="SBP Distribution of Men & Women", col="blue")
lines(kf, col="green")

I am completely new to all this! I'm pretty sure my exact question has also not been asked here but please lead me to any other resources that may help. Thanks.

Answer 1

The density objects have elements x and y elements that store the x-axis and distribution function values repectively. If you use the same from and to arguements for both density() calls, then the x values calculated should be the same.

Store the xy values for both densities in a dataframe an d then merge/join them on x , then you can calculate the difference and plot them:

x <- rnorm(1000,0,1)
y <- rnorm(1000,1,1)
fx <- density(x,from = -5,to=5)
fy <- density(y,from = -5,to=5)
plot(fx,col='blue',main="SBP Distribution of Men & Women")
lines(fy, col="green")

dfx <- data.frame(x=fx$x,
                  fx=fx$y)

dfy <- data.frame(x=fy$x,
                  fy=fy$y)

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(ggplot2)

inner_join(dfx,dfy,on='x') %>% 
  mutate(diff=fx-fy) %>% 
  ggplot()+
  geom_line(aes(x=x,y=diff))
#> Joining, by = "x"

^{Created on 2020-03-10 by the reprex package (v0.3.0)}