# [R] how to compute this summation...

Vitalie S. vitosmail at rambler.ru
Fri Aug 21 20:08:41 CEST 2009

```On Thu, 20 Aug 2009 02:36:38 +0200, kathie <kathryn.lord2000 at gmail.com>
wrote:

>
> Dear R users,
>
> I try to compute this summation,
>
> http://www.nabble.com/file/p25054272/dd.jpg
>
> where
>
> f(y|x) = Negative Binomial(y, mu=exp(x' beta), size=1/alp)
>
> http://www.nabble.com/file/p25054272/aa.jpg
>
> http://www.nabble.com/file/p25054272/cc.jpg
>
>
> In fact, I tried to use "do.call" function to compute each u(y,x) before
> the
> summation, but I got an error, "Error in X[i, ] * t(beta) :
> non-conformable
> arrays".
>
>
> Here is my code.
>
> #----------------------------------------------------------------------------------
>
>
>
> # data ------------------------------------------------------------
> yy <- c(2,2,10,3,6,7,9,9,14,6)
> x1 <- c(0,0,0,0,0,1,0,0,0,0)
> x2 <- c(1,1,0,1,0,0,0,0,0,0)
> x3 <- c(0,0,1,0,0,0,0,0,0,0)
> x4 <- c(0,0,0,0,0,0,0,0,0,0)
> x5 <- c(0,0,0,0,0,0,0,0,0,1)
> x6 <- c(0,0,0,0,1,0,0,0,0,0)
>
> n <- length(yy)
> x <- cbind(x1,x2,x3,x4,x5,x6)
> X <- cbind(1,x)
> p <- 7   ####  p:#beta
>
> library(numDeriv)
>
> # u function -------------------------------------------------------
>
> obj <- function(theta,i,j)
> {
> 	beta <- theta[1:p]
> 	alp <- theta[p+1]
>
> 	log(dnbinom(yy[j], mu=exp(rowSums(X[i,] * t(beta))), size=1/alp))
> }
>
>
>
> list <- expand.grid(j=seq(1,n),i=seq(1,n))
>
> func <- function(i,j) { grad(func=obj,i=i,j=j,
> x=c(1.50, 0.02, 0.09, 0.22,  0.17, -0.08, -0.13, 0.04)) }
>
> do.call( func, list )

I could not get what you were trying to do here, but here are a couple of
suggestions shat might help:

--Look at function ?outer (applies you function to combinations of i and j
- so no need of expand.grid)
--Look at ?Reduce - might help in your case.
--grad function does not have i and j variables - so you are doing
something wrong there
--do not name your data.frame - "list"
--do not call your function - "obj"

Best,
Vitalie.

> #------------------------------------------------------------------------------------
>
>
> Any suggestion will be greatly appreciated.
>
> Regards,
>
> Kathryn Lord

--

```