[R] Generalized cumsum?

[William Dunlap]

> -----Original Message-----
> From: r-help-bounces at r-project.org 
> [mailto:r-help-bounces at r-project.org] On Behalf Of OKB (not okblacke)
> Sent: Wednesday, September 16, 2009 6:04 PM
> To: r-help at stat.math.ethz.ch
> Subject: Re: [R] Generalized cumsum?
> 
> David Winsemius wrote:
> 
> >>         No, Reduce reduces an entire vector to a single value by  
> >>    repeatedly 
> >> combining adjacent elements.  I'm looking to convert a vector to  
> >> another vector where each element is some arbitrary aggregating
> >> function   applied to the first n elements of the vector. 
> > 
> > Yes. You need to look again:
> > 
> > accumulate=FALSE by default, but is subject to change.-
> 
>     	Ah, sorry.  Yes, you're right.  However, this still 
> doesn't satisfy 
> my needs, because since Reduce successively combines one element at a 
> time, it doesn't work for functions that don't self-compose 
> transitively.  For instance:
> 
> > Reduce(mean, c(1,2,3,4), accumulate=T)
> [1] 1 1 1 1
> 
> but I want
> 
> > cumapply(mean, c(1,2,3,4))
> [1] 1 1.5 2 2.5
> 
>     	Is there anything this general?

You can wrap mean() with a function that maintains sufficient
data that you can update the estimate.  I don't think this can
be automated, aside from the trivial method of retaining all
the data up until now so you can compute the statistic from
scratch each time.  Doesn't median require almost this approach?
(median could keep the accumulated data in sorted order
to make updates faster but it needs all the data.)  Most robust
techniques don't have update procedures because they depend
on order statistics of functions of the data.

Combining functions like seq, cumsum, cummax, and filter will
cover a lot of common cases.
   > cummean<-function(x)cumsum(x)/seq_along(x)
   > cummean(1:4)
   [1] 1.0 1.5 2.0 2.5

Bill Dunlap
TIBCO Software Inc - Spotfire Division
wdunlap tibco.com 
> 
> Thanks,
> -- 
> --OKB (not okblacke)
> Brendan Barnwell
> "Do not follow where the path may lead.  Go, instead, where there is
> no path, and leave a trail."
> 	--author unknown
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide 
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>