Use character input to define function in R

Question

I have an input data set, which may look something like this:

DF=data.frame(
    Variable = c("Test1", "Test2", "Test3"), 
    Distribution = c("Normal", "Exponential","Poisson"),
    Variable = c(2, 3, 4), 
    SD = c(2, NA, NA))

I want to use the random probability functions (e.g. rnorm rexp and rbinom) using the distributions given in the data frame DF.

So, how do I turn the text input into the correct functions?

I want to use the corresponding values in the Variable and SD columns as the mean values/standard deviations if appropriate.

Those functions have different input arguments that your DF doesn't have (eg., size, n). How do you think dealing with that? — pogibas
– pogibas, Commented Mar 14, 2018 at 11:21
It was just an example - the only missing parameter is n which for arguments sake say will be 10 for all distributions. The Variable column will be the mean for the Normal distribution, rate for Exponential, and lambda for Poisson. Should have everything. — sym246
– sym246, Commented Mar 14, 2018 at 11:23

bartektartanus · Accepted Answer · 2018-03-14 15:16:46Z

@r.user.05apr solution is working, but involves some expression parsing which is not needed here. We might get this a lot easier by creating a list of functions to use them later.

# generating data:
DF=data.frame(
  Variable = c("Test1", "Test2", "Test3"), 
  Distribution = c("Normal", "Exponential","Poisson"),
  VariablePrm = c(2, 3, 4), 
  SD = c(2, NA, NA),
  stringsAsFactors = FALSE)

# creating function list and selecting this functions by Distribution column
fun_vec <- c(Normal=rnorm, Exponential=rexp, Poisson=rpois)
DF$fun <- fun_vec[DF$Distribution]

# if SD is NA then simply call function only with variablePrm
# else call with sd
# 10 is the number of observations to generate
generate <- function(x) {
  if(is.na(x$SD)){
    x$fun(10, x$VariablePrm)
  }else{
    x$fun(10, x$VariablePrm, x$SD)
  }
}
# if we apply this functions to each row we will get matrix of results
# each column will have 10 rows of generated data for previously selected distribution
apply(DF, 1, generate)

r.user.05apr · Accepted Answer · 2018-03-14 13:35:32Z

0

Something like:

DF=data.frame(
  Variable = c("Test1", "Test2", "Test3"), 
  Distribution = c("Normal", "Exponential","Poisson"),
  VariablePrm = c(2, 3, 4), 
  SD = c(2, NA, NA),
  stringsAsFactors = FALSE)

# functions-lookup
fun_vec <- c("rnorm", "rexp", "rpois")
names(fun_vec) <- c("Normal", "Exponential", "Poisson")
DF$fun <- fun_vec[DF$Distribution]

# create expr
my_expr <- function(x) {
  txt <- paste0(x[1], "<-", x[5], "(", 10, ", ", x[3], 
                ifelse(is.na(x[4]), "", 
                       paste0(", ", x[4])), ")")
}
want <- apply(DF, 1, function(x) eval(parse(text = my_expr(x))))
colnames(want) <- DF$Variable
want

answered Mar 14, 2018 at 13:35

r.user.05apr

5,4783 gold badges27 silver badges43 bronze badges

Collectives™ on Stack Overflow

Use character input to define function in R

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related