Re: about input parameters

[Sundar Dorai-Raj <sundar.dorai-raj@pdf.com>]

Stefano Sofia wrote:
> Dear Splus users,
> I have got some problems about input parameters for a Splus personal
> procedure.
>
> If I call the procedure below reported 
> myprocedure(-38.94376,0.001842975)
> there is an error in "smallg(Y1)" which says: 
> Error in small(Y1): Object "E7" not found.
> Why? 
> E7 and V7 are two scalars.
> Defining E7 and V7 as 
> array(NA, dim = c(1, 1)),
> mode(E7) is "numeric", mode(V7) is "numeric". 
> I encounter the same problem if I define E7 and V7 as matrices of dimesion 1
> by 1 
> or even if I do not define them at all.
> Why smallg can't be evaluated for eta=E7?
> If I use the internal procedure without E7 and V7 but with pure numbers, i.e.
>
> internal(-38.94376,0.001842975,1)
> everything works fine. Why?
>
> I am using Splus4 with W XP.
>
> grateful for your help
> Stefano

This is a scoping issue. Replace function internal with:

  internal <- function(normalmean, normalvariance, state){
    alpha <- 0.01
    if(state == 0) {
      smallg <- substitute(function(eta, mu, sigma){
        (dnorm(eta, mean = mu, sd = sigma) * 1)/(1 + exp(alpha * eta))})
    } else {
      smallg <- substitute(function(eta, mu, sigma){
        (dnorm(eta, mean = mu, sd = sigma) * exp(alpha * eta))/(1 + 
exp(alpha * eta))})
    }
    repeat {
      # STEP 1: SAMPLE A POINT Y1 FROM bigg(Y)
      Y1 <- rnorm(1, mean = normalmean, sd = sqrt(normalvariance))
      # STEP 2: SAMPLE A UNIFORM(0,1) RANDOM VARIABLE U
      U <- runif(1, min = 0, max = 1)
      # STEP 3: IF U <= g(Y1)/G(Y1) ACCEPT Y1;
      # REPEAT THIS LOOP UNTIL ONE Y IS ACCEPTED
      Y1.num <- smallg(Y1, normalmean, sqrt(normalvariance))
      Y1.den <- dnorm(Y1, mean = normalmean, sd = sqrt(normalvariance))
      if(U <= Y1.num/Y1.den) {
        sampledvalue <- Y1
        break
      }
    }
    result <- sampledvalue
    result
  }