How to call a function using *apply with different data, where output is list and intput is list of 2x2 matrix?

Question

I have data from 10 runs of a simulation, stored as a list. I want to call a function FUN1 with this data, without repeating the code 10 times. The output from FUN1 is a value of model parameter such that:

               $theta=5
               $Theta= 0.5
               $pi_1 = 0.6
               $pi_2 = 0.4 
               $loglik_1 = 123.6
               $loglik_2 = 23.56

So, this how the output of FUN1 when I fit it to one of the 10 runs. I would like to have the output of all runs without refit my function for each run.

I know that we can use loop-family functions in R such as lapply (list) or tapply (vector) functions. I tried both of them but both did not work.

Here is my data:

   library(VineCopula)
   library(copula)
   Runs = 10
   Saveas = vector(mode = "list", length = Runs)
   pb <- txtProgressBar(min = 0, max = Runs, style = 3)

   for(j in 1:Runs) {
     setTxtProgressBar(pb, j)
     N=2000
     dim = dim
     U = runif(N, min=0,max=1)
     X = matrix(NA, nrow=N, ncol=2)

     inds <- U < 0.7

     X[inds, ]  <- rCopula(sum(inds),
                    claytonCopula(1, dim=2))

     X[!inds, ] <- rCopula(N - sum(inds),
                    frankCopula(4, dim=2))
     Saveas[[j]] = X
   }

This is my function:

      FUN1 <- EM_mixture_copula(data =   
          Saveas[[j]],pi_1=pi_1,pi_2=pi_2,theta = theta, 
          Theta=Theta, tol = .00001,     maxit = 1000)

Here is my tries with the errors that I got:

    > result <- tapply(X,FUN1,simplify = T)
    Error in tapply(X, FUN, simplify = T) : arguments must have same length.

    > Result <– lapply(X,FUN1)
    Error in get(as.character(FUN), mode = "function", envir = envir) : object 'F' of mode 'function' was not found.

Note that the output of copula is a matrix(nrow=N, ncol=2) (since the dimension of copula is 2). For example:

   xx <-            

        rCopula(N=4 ,claytonCopula(0.5))
            xx
                   [,1]        [,2]
      [1,] 0.6269311043 0.229429156
      [2,] 0.3257583519 0.268244546
      [3,] 0.7446442267 0.436335203
      [4,] 0.3186246504 0.163209827

Where 0.5 is copula parameter.

Any help, please?

X is my data (The data is repeated 10 times).all the runs of 10 simulation was saved as list. — F.family, Mar 10 '17 at 23:55
I thought your data was in `Saveas`. You need to define a function to and use it with lapply. But `FUN1` is not a function. — DJJ, Mar 10 '17 at 23:59
I guess you mean something like `FUN1 <- function(j){ EM_mixture_copula(data = Saveas[[j]],pi_1=pi_1,pi_2=pi_2,theta = theta, Theta=Theta, tol = .00001, maxit = 1000) }` — DJJ, Mar 11 '17 at 00:02
FUN1 is a function to fit EM algorithm to my data. But instead to run this function 10 times using the data Saveas[[1]] and Saveas[[2]], etc. I would like to repeat my function to all the 10 simulation run in a convenient way. — F.family, Mar 11 '17 at 00:02
I try for loop, but did not get the result also I did not get error. I will try your guess and see. — F.family, Mar 11 '17 at 00:05
Most of us don't have libraries `copula` and `VineCopula` installed so your question shouldn't require us to since they're not strictly necessary; just fake them out with random-seeded data generated from `set.seed(...)`, `rnorm()` calls, `as.matrix()` and whatever else. ("Minimal Reproducible Example") — smci, Mar 11 '17 at 01:57
Thank you for your edit and comment. The lapply and tapply or other function work well with any other data or function. It just did not work with my case. My data generated from copula is as a matrix. However, the out but of 'FUN1' is not a matrix. It the value of theta, Theta (copula parameters) and pi_1 and pi_2 ( the mixture coefficients). So, I would like to have these values 10 times for the 10 run simulation data without repeating FUN1 10 times. — F.family, Mar 11 '17 at 20:19

Sathish · Answer 1 · 2017-03-11T01:04:17.473

I guess dim inside sim_fun should be 1. Since, I do not have data or the libraries installed, this answer is an unverified post. However, it will help you figure out the solution from here.

sim_fun <- function( N )
{
  U=runif(N, min=0,max=1)
  inds <- U < 0.7
  X <- matrix(NA, nrow = N, ncol = 2)
  X[inds, 1:2] <- rCopula(sum(inds), claytonCopula(1, dim=2))
  X[!inds, 1:2] <- rCopula(N - sum(inds), frankCopula(4,dim=2))
  return( X )
}

set.seed(1L)  # set state of random number generator
sim_data <- replicate( n = 10, sim_fun( N = 2000 ))  # get simulated data 10 times

# apply EM function on the simulated data
apply( sim_data, 3, function( x ) EM_mixture_copula(data =  x,
                                                    pi_1 = pi_1,
                                                    pi_2=pi_2,
                                                    theta = theta, 
                                                    Theta=Theta, 
                                                    tol = .00001,
                                                    maxit = 1000))

How to call a function using *apply with different data, where output is list and intput is list of 2x2 matrix?

1 Answers1