12

I fitted the normal distribution with fitdist function from fitdistrplus package. Using denscomp, qqcomp, cdfcomp and ppcomp we can plot histogram against fitted density functions, theoretical quantiles against empirical ones, the empirical cumulative distribution against fitted distribution functions, and theoretical probabilities against empirical ones respectively as given below.

set.seed(12345)
df <- rnorm(n=10, mean = 0, sd =1)
library(fitdistrplus)
fm1 <-fitdist(data = df, distr = "norm")
summary(fm1)

denscomp(ft = fm1, legendtext = "Normal")

enter image description here

qqcomp(ft = fm1, legendtext = "Normal")

enter image description here

cdfcomp(ft = fm1, legendtext = "Normal")

enter image description here

ppcomp(ft = fm1, legendtext = "Normal")

enter image description here

I'm keenly interested to make these fitdist plots with ggplot2. MWE is below:

qplot(df, geom = 'blank') +
  geom_line(aes(y = ..density.., colour = 'Empirical'), stat = 'density') +  
  geom_histogram(aes(y = ..density..), fill = 'gray90', colour = 'gray40') +
  geom_line(stat = 'function', fun = dnorm, 
            args = as.list(fm1$estimate), aes(colour = 'Normal')) +
  scale_colour_manual(name = 'Density', values = c('red', 'blue'))

enter image description here

ggplot(data=df, aes(sample = df)) + stat_qq(dist = "norm", dparam = fm1$estimate)

How can I start making these fitdist plots with ggplot2?

halfer
  • 19,824
  • 17
  • 99
  • 186
MYaseen208
  • 22,666
  • 37
  • 165
  • 309
  • 3
    If this hadn't a bounty attached I would vote to close as too broad. Each graph should be a different question (although you might not need to ask for each graph if you had an answer for one or two of them). – Roland Jul 30 '15 at 08:50

1 Answers1

4

you could use something like that:

library(ggplot2)

ggplot(dataset, aes(x=variable)) +
geom_histogram(aes(y=..density..),binwidth=.5, colour="black", fill="white") +
stat_function(fun=dnorm, args=list(mean=mean(z), sd=sd(z)), aes(colour =
"gaussian", linetype = "gaussian")) + 
stat_function(fun=dfun, aes(colour = "laplace", linetype = "laplace")) + 
scale_colour_manual('',values=c("gaussian"="red", "laplace"="blue"))+
scale_linetype_manual('',values=c("gaussian"=1,"laplace"=1))

you just need to define dfun before running the graphic. In this example, it's a Laplace distribution but you can pick any you want and add some more stat_function if you want.

MLavoie
  • 9,671
  • 41
  • 36
  • 56