How to generate a negative binomial distribution with different sample sizes for power analyses in R-CodePudding

I try to do a power simulation with an outcome variable that is zero-inflated. So I use a negative binomial distribution. What I need is the following distribution:

library(tidyverse)
set.seed(123)
rt_random <- 
rnbinom(n = 2000, mu = 25, size = .9) 
qplot(rt_random)

desired distribution

Now, I tried to create a function where I can use different sample sizes for the power simulation. I use the pmapfunction for this. However this does not seem to work. The final distribution is not zero-inflated at all and the mean is not close to the defined mean:

generate_distribution <- function(n, mus, sizes){
  tibble(n = n, 
       t_mu = mus, 
       t_size = sizes)%>% 
  mutate(rt = pmap(list(n, t_mu, t_size),
                     ~rnbinom(..1, ..2, ..3))) %>% 
  unnest(rt) 
  
}
set.seed(123)
rt_df <- 
  generate_distribution(n = 2000,
                        mus = 25,
                        sizes = .9)
qplot(rt_df$rt)

wrong distribution

Is there an easy way to change my code so that I get the desired distribution?

CodePudding user response：

The problem is that in your lambda function, you are not using the argument names for rnbinom. The default order for the arguments if you don't name them is n, size, prob, mu, so you are passing 2000 to n, 25 to size and 0.9 to prob. Just name the arguments explicitly as you did in your first example, and your code will work.

generate_distribution <- function(n, mus, sizes){
  tibble(n = n, 
       t_mu = mus, 
       t_size = sizes)%>% 
  mutate(rt = pmap(list(n, t_mu, t_size),
                     ~rnbinom(n = ..1, mu = ..2, size = ..3))) %>% 
  unnest(rt) 
}

set.seed(123)
rt_df <- 
  generate_distribution(n = 2000,
                        mus = 25,
                        sizes = .9)
qplot(rt_df$rt)