Fitdist (package fitdistrplus) with ggplot2

tidyverse
1

I need a help to insert the fitdist into ggplot2. I tried to find an answer on search sites.

The basic code I find whitout ggplot is:

library(fitdistrplus)

data = c(1, 2, 3, 5)

gama_distribution = fitdist(data, distr = 'gamma', method = 'mle')
normal_distribution = fitdist(data, distr = 'norm', method = 'mle')
lognormal_distribution = fitdist(data, distr = 'lnorm', method = 'mle')

plot(gama_distribution, demp = TRUE)
plot(normal_distribution, demp = TRUE)
plot(lognormal_distribution, demp = TRUE)

I need a help to combine fitdist with ggplot2.

2

Someone knows how to solve this problem?

3

I need help to making fitdist into a graphic with ggplot!

4

(post deleted by author)

5

Look at the documentation for the package. You will see several functions, such as cdfcomp, that allow an argument plotstyle="ggplot" and will then use ggplot.

6

I try using the code below, but the dataframe, in data_1, don't accept the vector.

library(fitdistrplus)
library(ggplot2)

data = c(1, 2, 3, 5)

gama_distribution = fitdist(data, distr = 'gamma', method = 'mle')

data_1 = data.frame(gama_distribution)

data_2 = ggplot(data_1)

So, I don't know how to integrate fitdist with ggplot.

7

The problem is that the result of fitdist isn't a dataframe or anything that can be turned into a dataframe (or tibble). You need to either create a dataframe yourself from the contents of the fitdist function or to use the functions like cdfcomp that will do the translation internally and create a ggplot object.

8

library(fitdistrplus)
library(ggplot2)

data = c(1, 2, 3, 5)

data_1 = data.frame(fitdist(data, distr = 'gamma', method = 'mle'))

data_2 = ggplot(data_1)

I try with that code above, but still don't work (don't accept the vector).

9

Really. See the reply above.

10

Well, I just give up to solve this problem with myself, for a while, because I'm not a programmer. I just ask for help to someone show me the solution with a code, for me to apply a biological analysis.

11

According to the fistdistrplus Vignette, you can return a ggplot using the package's plotting functions and the plotstyle argument. For example:

p = denscomp(list(gama_distribution, normal_distribution, lognormal_distribution),
             plotstyle="ggplot")

Rplot
Rplot674×544 25.3 KB

See the help for denscomp, which lists other types of plots you can create.

1 Like
12

Here's a more complete example that shows you how to calculate the density values for each estimated distribution and put them in a data frame. Then you can create your own plot and structure it exactly how you want.

library(fitdistrplus)
#> Loading required package: MASS
#> Loading required package: survival
library(tidyverse)
theme_set(theme_bw())

# Fake data
set.seed(3)
data = rgamma(1000, 6.6, 2.3)

# Create distributions
dists = list(Gamma="gamma", Normal="norm", Lognormal="lnorm") %>% 
  map(~fitdist(data, distr=.x, method="mle"))

# Look at structure of one of the distribution objects so we know where to 
# get its density function name (the "distname") and the values for the 
# location and scale parameters (the "estimate")
str(dists[["Gamma"]])
#> List of 17
#>  $ estimate   : Named num [1:2] 6.78 2.39
#>   ..- attr(*, "names")= chr [1:2] "shape" "rate"
#>  $ method     : chr "mle"
#>  $ sd         : Named num [1:2] 0.296 0.108
#>   ..- attr(*, "names")= chr [1:2] "shape" "rate"
#>  $ cor        : num [1:2, 1:2] 1 0.963 0.963 1
#>   ..- attr(*, "dimnames")=List of 2
#>   .. ..$ : chr [1:2] "shape" "rate"
#>   .. ..$ : chr [1:2] "shape" "rate"
#>  $ vcov       : num [1:2, 1:2] 0.0876 0.0309 0.0309 0.0117
#>   ..- attr(*, "dimnames")=List of 2
#>   .. ..$ : chr [1:2] "shape" "rate"
#>   .. ..$ : chr [1:2] "shape" "rate"
#>  $ loglik     : num -1453
#>  $ aic        : num 2910
#>  $ bic        : num 2920
#>  $ n          : int 1000
#>  $ data       : num [1:1000] 1.72 2.19 1.56 3.02 2.67 ...
#>  $ distname   : chr "gamma"
#>  $ fix.arg    : NULL
#>  $ fix.arg.fun: NULL
#>  $ dots       : NULL
#>  $ convergence: int 0
#>  $ discrete   : logi FALSE
#>  $ weights    : NULL
#>  - attr(*, "class")= chr "fitdist"

# Vector of values at which we want to get density
xvals = seq(min(data), max(data), length.out=100)

# Use each distribution's location and scale parameters to create a data frame of 
#  values to plot
dist.vals = map_df(dists,
                   ~{
                     # Get desired density function
                     fun = paste0("d", .x$distname)
                     fun = match.fun(fun)
                     
                     # Get density values for that function
                     fun(xvals, .x$estimate[1], .x$estimate[2])
                   })

dist.vals = data.frame(xvals, dist.vals) %>% 
  pivot_longer(-xvals, names_to="Distribution", values_to="Density")

dist.vals %>% 
  ggplot() +
  geom_histogram(data=data.frame(x=data), aes(x=x, y=after_stat(density)),
                 breaks=seq(floor(min(data)), ceiling(max(data)), 0.5), 
                 colour="white", fill="grey80",
                 ) +
  geom_line(aes(xvals, Density, colour=Distribution)) +
  scale_x_continuous(breaks=0:10)

Created on 2025-04-04 with reprex v2.1.1

13

Alright. Thanks for tour help!