Re: [S] is there a cumouter() like function?

Prof Brian D Ripley (ripley@stats.ox.ac.uk)
Tue, 25 Aug 1998 07:37:42 +0100 (BST)


On Mon, 24 Aug 1998, Jack Lewis wrote:

> Oops, what I meant in the example below was:
> Given a sequence a[i], i=0,1,2,...,n, and some constant k,
> with 0 < k < 1, compute the sequence
> b[i] = k*b[i-1] + a[i], i=1,2,...,n
> NOT
> b[i]= k*a[i-1]+a[i]
> Sorry, Jack
> P.S. I know it can be done with apply()

Um. Your final example is, I think,

X[i] = k*X[i-i] + a[i], i=1,2,...,n

and this cannot be done directly with apply. Actually, it cannot be done at
all unless we know X[0], so let us assume this is specified. It is then an
AR(1) process that can be solved analytically:

X[1] = k*X[0] + a[1]
X[2] = k*X[1] + a[2] = k*k*X[0] + k*a[1] + a[2]
X[3] = k^3*X[0] + k^2*a[1] + k*a[2] + a[3]
...
so X = [k^i]X[0] + [k^(i-j-1)I(i>=j)]a

and that can be done by matrix multiplication in S. Alternatively, the
system of equations is

X[0] = given
X[i] - k*X[i-1] = a[i]

and that is of the form A %*% X = a and can be solved by solve. As A is so
simple, solve is overkill (even using the forms in library Matrix).

For an example of a calculation that cannot be done by a loop or
analytically, see V&R2 pp.116-7.

>
> Jack Lewis wrote:
>
> > Here is a simple example that I think requires looping,
> > (although I'd be thrilled to find out I'm wrong).
> > Given a sequence a[i], i=0,1,2,...,n, and some constant k,
> > with 0 < k < 1, compute the sequence
> > b[i] = k*a[i-1] + a[i], i=1,2,...,n

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu. To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message: unsubscribe s-news