right/left skewed normal distribution curve in R and shading the area?

How can we create skewed normal distribution curve in R ?

My aim is to produce skewed normal distribution to represent accounting CVA (expcted losses) and regulatory CVA (99% quantile). I want to have these in a same figure and shaded

These are the codes that I have used, to produce the figure. However, when i introduce shade function, the skew despair.

Can anyone please help

library(sn)

X <- seq(-1, 2, 0.01)
plot(X, dsn(X, xi = 0.1, omega = 0.3, alpha = 5), type = "l")
abline(v = 0.2)
shadenorm()
shadenorm(below=1, justbelow = TRUE, color = 'blue')

I would also like to label axis, y labelled as probability while x labelled exposure.

I know that there is similar post [quote="AbhishekHP, post:1, topic:39115, full:true"]
How can we create skewed normal distribution curve in R ?
How to vary the skewedness using a variable ?
i.e. height of the peak and tail of the plot ?

Although many links say that they have an answer but non worked



This is in addition to earlier question:
How to calculate the cumulative area of the overlapped normal distribution curve in R ? solved by @Yarnabrina

# Tried 
library(fGarch)
library(tidyverse)
N <- 10000
x <- rnbinom(N, 10, .5)
dsnorm(x, mean = 0, sd = 1, xi = 1.5, log = FALSE)
# psnorm(x, mean = 0, sd = 1, xi = 1.5) %>% plot()
# qsnorm(x, mean = 0, sd = 1, xi = 1.5) %>% plot()
# rsnorm(x, mean = 0, sd = 1, xi = 1.5) %>% plot()

plot(dsnorm(x, mean = 0, sd = 1, xi = 1.5, log = FALSE))
[/quote]

but it does not answer my question

I do not understand what is the problem. My previous code seems to work fine, if I guess correctly what you want to do. I can not be sure, as I do not know what does shadenorm do, or which package contains this function. Here's an example with base graphics, and you can switch to ggplot2 if you prefer.

library(sn)
#> Loading required package: stats4
#> 
#> Attaching package: 'sn'
#> The following object is masked from 'package:stats':
#> 
#>     sd

values <- seq(from = -1,
              to = 2,
              by = 0.01)
densities <- dsn(x = values,
                 xi = 0.1,
                 omega = 0.3,
                 alpha = 5)

plot(x = values,
     y = densities,
     type = "l")
abline(v = 0.2)
polygon(x = c(min(values), values[values < 0.2], 0.2),
        y = c(0, densities[values < 0.2], 0),
        border = NA,
        col = adjustcolor(col = "blue",
                          alpha.f = 0.5))

If this does not help, please provide a reproducible example. And please take a look here:

1 Like

Thanks @Yarnabrina the codes are working fine, however, what if I want to add another shading in that same graph, say to indicate 99% percentile. The codes are working fine.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

This is a general solution to shade below a point, above a point and in between two points. I have intentionally didn't use mean and quantile so that you can do that yourself. Hope this helps.

# loading library
library(sn)
#> Loading required package: stats4
#> 
#> Attaching package: 'sn'
#> The following object is masked from 'package:stats':
#> 
#>     sd

# generating data
exposures <- seq(from = -1,
                 to = 2,
                 by = 0.01)
probabilities <- dsn(x = exposures,
                     xi = 0.1,
                     omega = 0.3,
                     alpha = 5)

# calculating measures
measures <- quantile(x = exposures,
                     probs = c(0.40, 0.60))

# plotting distribution
plot(x = exposures,
     y = probabilities,
     type = "l")

# adding vertical lines at selected measures
abline(v = c(measures[1], measures[2]))

# shading areas below arithmetic mean, above upper 0.01 point and in between
polygon(x = c(min(exposures), exposures[exposures <= measures[1]], measures[1]),
        y = c(0, probabilities[exposures <= measures[1]], 0),
        border = NA,
        col = adjustcolor(col = "red",
                          alpha.f = 0.5))
polygon(x = c(measures[1], exposures[(exposures >= measures[1]) & (exposures <= measures[2])], measures[2]),
        y = c(0, probabilities[(exposures >= measures[1]) & (exposures <= measures[2])], 0),
        border = NA,
        col = adjustcolor(col = "green",
                          alpha.f = 0.5))
polygon(x = c(measures[2], exposures[exposures >= measures[2]], max(exposures)),
        y = c(0, probabilities[exposures >= measures[2]], 0),
        border = NA,
        col = adjustcolor(col = "blue",
                          alpha.f = 0.5))

Created on 2020-04-11 by the reprex package (v0.3.0)

1 Like

I manage to plot using the following codes however I am failing to cut the (abline = mu) and (abline = quan) to stop at the distribution curve only

library(sn)
#> Loading required package: stats4
#> 
#> Attaching package: 'sn'
#> The following object is masked from 'package:stats':
#> 
#>     sd

Exposures <- seq(from = -2,
              to = 4,
              by = 0.01)
Probability <- dsn(x = Exposures,
                 xi = 0.1,
                 omega = 0.72,
                 alpha = 5)
plot(x = Exposures,
     y = Probability,
     main = "probability distribution of exposure, representing accounting CVA",
     type = "l")
mu=mean(Exposures)
quan=quantile(Exposures,c(0.99))
abline(v = quan)
abline(v = mu)
polygon(x = c(min(Exposures), Exposures[Exposures < mu], mu),
        y = c(0, Probability[Exposures < mu], 0),
        border = NA,
        col = adjustcolor(col = "blue",
                          alpha.f = 0.5)
        )

I have also tried to shade the area between the mean (mu) and quan with the following codes but seems not to be working properly.

polygon(x = c(mu,Exposures[Exposures < quan], mu),
        y = c(0, Probability[Exposures < quan], 0),
        border = NA,
        col = adjustcolor(col = "red",
                          alpha.f = 0.5))

Help please