[R] Yearly aggregates and matrices

[Gabor Grothendieck]

On Wed, Apr 20, 2011 at 5:49 AM, mathijsdevaan <mathijsdevaan at gmail.com> wrote:
> As a follow up on this post, I am trying to slightly adjust the solution
> kindly provided by Gabor. However, I am getting some results that I do not
> understand. Example:
>
> # devel version of zoo
> install.packages("zoo", repos = "http://r-forge.r-project.org")
> library(zoo)
>
> DF1 = data.frame(read.table(textConnection("    B  C  D  E  F  G
> 8025  1995  0  4  1  2
> 8025  1997  1  1  3  4
> 8026  1995  0  7  0  0
> 8026  1996  1  2  3  0
> 8026  1997  1  2  3  1
> 8026  1998  6  0  0  4
> 8026  1999  3  7  0  3
> 8027  1997  1  2  3  9
> 8027  1998  1  2  3  1
> 8027  1999  6  0  0  2
> 8028  1999  3  7  0  0
> 8029  1995  0  2  3  3
> 8029  1998  1  2  3  2
> 8029  1999  6  0  0  1"),head=TRUE,stringsAsFactors=FALSE))
>
> a <- read.zoo(DF1, split = 1, index = 2, FUN = identity)
> sum.na <- function(x) if (any(!is.na(x))) sum(x, na.rm = TRUE) else NA
> b <- rollapply(a, 3,  sum.na, align = "right", partial = TRUE)
> newDF <- lapply(1:nrow(b), function(i)
>       prop.table(na.omit(matrix(b[i,], nc = 4, byrow = TRUE,
>               dimnames = list(unique(DF1$B), names(DF1)[-1:-2]))), 1))
> names(newDF) <- time(a)
> c<-lapply(newDF, function(mat) tcrossprod(mat / sqrt(rowSums(mat^2))))
>
> Now I would like the elements e in c to be equal to 1-e. However,
>
> c<-lapply(newDF, function(mat) 1 - tcrossprod(mat / sqrt(rowSums(mat^2))))
>
> gives a value  of 2.220446e-16 for as.data.frame(c['1999'])[2,2] instead of
> 0
>
> What am I doing wrong here? Thanks a lot!
>

See FAQ 7.31 at:

http://cran.r-project.org/doc/FAQ/R-FAQ.html#Why-doesn_0027t-R-think-these-numbers-are-equal_003f


-- 
Statistics & Software Consulting
GKX Group, GKX Associates Inc.
tel: 1-877-GKX-GROUP
email: ggrothendieck at gmail.com