creating a function

Hi,
thank you Rui for your reply.
I understand I haven?t made it clear enough.
What I do want to make is a kind of autoregressive process where Y_t+1 depends on its lagged value Y_t.
If I wanted to simulate a simple AR(1) process I would have done something as follows:

n<-500
Y_init_cond<-0
Et<-ts(rnorm(n,mean=0,sd=1))
Yt<-Et*0
Yt[1]<-Y_init_cond+Et[1]
for(i in 2:(n)) {
   Yt[i]<-Yt[i-1]+Et[i]
}
Yt<-ts(Yt[(length(Yt)-n+1):length(Yt)])

where Et is a vector of random errors.
Since I do not want to simulate a simple autoregressive process the code above is no longer suitable for my purpose.
In fact what I want to do is a function as Y_t+1 = Y_t + (1\p)*summation(x(Y_t,Y_t-1,...)) + epsilon.t,
where Y_t+1 depends on its lagged value Y_t, but also on 1/p (which is a parameter) times summation(x(Y_t,Y_t-1,...)) where x is a function that also depends on the lagged values of Y_t+1 from Y_t to Y_t-n. Epsilon_t is, as You said a vector of random errors.

Any suggestion is very appreciated.

thanks and best regards,
Simone

From: Rui Barradas
Sent: Saturday, December 22, 2012 8:29 PM
To: Simone Gogna
Cc: r-help at r-project.org
Subject: Re: [R] creating a function

Hello,

It's a bad idea to name a function and one of it's arguments Y, use Yfun and Y.
What does summation(x(Y.t, Y.t-1, ...)) mean? Is there a multiplication sign between x and (Y.t, ...)?
And is epsilon a vector of errors, one for each Y.t?
If so, the following might do it.


Yfun <- function(Y, p, x, epsilon){
     for (i in 2:length(Y)) {
         Y[i] <- Y[i-1] + (1/p)*sum(x*Y[1:(i-1)]) + epsilon[i]
     }
     Y
}


Hope this helps,

Rui Barradas

Em 22-12-2012 09:29, Simone Gogna escreveu:
Dear R users,
I???d like to create a function as:

Y.t+1 = Y.t + (1\p)*summation(x(Y.t,Y.t-1,...)) + epsilon.t

where x is a function of Y.t, Y.t-1 and so on, epsilon is a random error and p is a parameter.

Do you think something of the following form might be appropriate?

Y<-function(Y,p,x,epsilon){
                        for (i in 2:length(Y)) {
                          Y[i]<-Y[i-1]+(1/p)*sum(x(Y[i-1]))+epsilo.t}
                        Y
}

Any indication is warmly appreciated.

thanks and best regards
	[[alternative HTML version deleted]]


    

Hi, 
thank you Rui for your reply.
I understand I haven???t made it clear enough.
What I do want to make is a kind of autoregressive process where Y_t+1
depends on its lagged value Y_t.
If I wanted to simulate a simple AR(1) process I would have done
something as follows:

n<-500
Y_init_cond<-0
Et<-ts(rnorm(n,mean=0,sd=1))
Yt<-Et*0
Yt[1]<-Y_init_cond+Et[1]
for(i in 2:(n)) {
 Yt[i]<-Yt[i-1]+Et[i]
}
Yt<-ts(Yt[(length(Yt)-n+1):length(Yt)])

where Et is a vector of random errors.
Since I do not want to simulate a simple autoregressive process the
code above is no longer suitable for my purpose.
In fact what I want to do is a function as Y_t+1 = Y_t +
(1\p)*summation(x(Y_t,Y_t-1,...)) + epsilon.t,
where Y_t+1 depends on its lagged value Y_t, but also on 1/p (which is
a parameter) times summation(x(Y_t,Y_t-1,...)) where x is a function
that also depends on the lagged values of Y_t+1 from Y_t to Y_t-n.
Epsilon_t is, as You said a vector of random errors.

Any suggestion is very appreciated.

thanks and best regards,
Simone

From: Rui Barradas 
Sent: Saturday, December 22, 2012 8:29 PM
To: Simone Gogna 
Cc: r-help at r-project.org 
Subject: Re: [R] creating a function

Hello,

It's a bad idea to name a function and one of it's arguments Y, use
Yfun and Y.
What does summation(x(Y.t, Y.t-1, ...)) mean? Is there a multiplication
sign between x and (Y.t, ...)?
And is epsilon a vector of errors, one for each Y.t?
If so, the following might do it.


Yfun <- function(Y, p, x, epsilon){
   for (i in 2:length(Y)) {
       Y[i] <- Y[i-1] + (1/p)*sum(x*Y[1:(i-1)]) + epsilon[i]
   }
   Y                      
}


Hope this helps,

Rui Barradas

Em 22-12-2012 09:29, Simone Gogna escreveu:
Dear R users,
I????????d like to create a function as:

Y.t+1 = Y.t + (1\p)*summation(x(Y.t,Y.t-1,...)) + epsilon.t

where x is a function of Y.t, Y.t-1 and so on, epsilon is a random
error and p is a parameter.

Do you think something of the following form might be appropriate?

Y<-function(Y,p,x,epsilon){
                      for (i in 2:length(Y)) {
                        Y[i]<-Y[i-1]+(1/p)*sum(x(Y[i-1]))+epsilo.t}
                      Y                      
}

Any indication is warmly appreciated.

thanks and best regards
[[alternative HTML version deleted]]