similar to: code for sum function

The algorithm does make a differece. You can use Kahan's summation algorithm (https://en.wikipedia.org/wiki/Kahan_summation_algorithm) to reduce the error compared to the naive summation algorithm. E.g., in R code: naiveSum <- function(x) { s <- 0.0 for(xi in x) s <- s + xi s } kahanSum <- function (x) { s <- 0.0 c <- 0.0 # running compensation for lost

Dear Tomas, Where do I find these files? Do they contain the code for the sum function? What do you mean exactly with your point on long doubles? Where can I find documentation on this? Cheers, Rampal On Mon, Feb 18, 2019, 15:38 Tomas Kalibera <tomas.kalibera at gmail.com wrote: > See do_summary() in summary.c, rsum() for doubles. R uses long double > type as accumulator on systems

Someone said it used a possibly platform-dependent higher-than-double-precision type. By the way, in my example involving rep(1/3, n) I neglected to include the most precise way to calculate the sum: n%/%3 + (n%%3)/3. Bill Dunlap TIBCO Software wdunlap tibco.com On Wed, Feb 20, 2019 at 2:45 PM Rampal Etienne <rampaletienne at gmail.com> wrote: > Dear Will, > > This is exactly

Dear Will, This is exactly what I find. My point is thus that the sum function in R is not a naive sum nor a Kahansum (in all cases), but what algorithm is it using then? Cheers, Rampal On Tue, Feb 19, 2019, 19:08 William Dunlap <wdunlap at tibco.com wrote: > The algorithm does make a differece. You can use Kahan's summation > algorithm

Specifically: https://svn.r-project.org/R/trunk/src/main/summary.c And if you don't want to deal with Subversion, you can look at the read-only github mirror: https://github.com/wch/r-source/blob/e5b21d0397c607883ff25cca379687b86933d730/src/main/summary.c#L115-L131 On Thu, Feb 21, 2019 at 11:57 AM David Winsemius <dwinsemius at comcast.net> wrote: > > > On 2/20/19 2:55 PM,

This SO question may be of interest: https://stackoverflow.com/questions/38589705/difference-between-rs-sum-and-armadillos-accu/ which points out that sum() isn't doing anything fancy *except* using extended-precision registers when available. (Using Kahan's algorithm does come at a computational cost ...) On 2019-02-19 2:08 p.m., William Dunlap via R-devel wrote: > The algorithm

Dear Rampal, you can download R source code in form of a tarball or from subversion, please see https://cran.r-project.org/doc/manuals/R-admin.html#Obtaining-R https://cran.r-project.org/doc/manuals/R-admin.html#Using-Subversion-and-rsync There is also a web access to subversion, so specifically the sum is available in https://svn.r-project.org/R/trunk/src/main/summary.c The definition of

On 2/20/19 2:55 PM, Rampal Etienne wrote: > Dear Tomas, > > Where do I find these files? Do they contain the code for the sum function? Yes. https://svn.r-project.org/R/trunk/ David > > What do you mean exactly with your point on long doubles? Where can I find > documentation on this? > > Cheers, Rampal > > On Mon, Feb 18, 2019, 15:38 Tomas Kalibera

(I didn't see anyone else answer this, so ...) You can probably find the R code in src/main/ but I'm not sure. You are talking about a very simple calculation, so it seems unlike that the algorithm is the cause of the difference. I have done much more complicated things and usually get machine precision comparisons. There are four possibilities I can think of that could cause (small)

Hello, I have the following data frame (DF): V5 V5.1 V5.2 V5.3 V5.4 V5.5 2 -5890.18905 -6019.84665 -6211.06545 -6198.9353 -6616.8677 -6498.7183 3 -5890.18905 -6019.84665 -6211.06545 -6198.9353 -6616.8677 -6498.7183 4 -5890.18905 -6019.84665 -6211.06545 -6198.9353 -6616.8677 -6498.7183 5 -5890.18905 -6019.84665 -6211.06545 -6198.9353 -6616.8677

Looking for more ways to speed up R, I've found that large improvements are possible in the speed of "sum" and "prod" for long real vectors. Here is a little test with R version 2.11.1 on an Intel Linux system > a <- seq(0,1,length=1000) > system.time({for (i in 1:1000000) b <- sum(a)}) user system elapsed 4.800 0.010 4.817 > system.time({for (i

Hi, I have set up KVM on FC12. I was wondering if its possible to Display the VM using spice + libvirt? As in, if I can define my domain with graphics like: <graphics type='spice' port='-1' autoport='yes' listen='0.0.0.0'/> or <graphics type='spice' port='5903' '

The replay gain code has dos line endings in CVS, which causes problems for the Sun compiler, among others. Attached is a patch for the lazy, but it's probably easier to fix locally and commit. -r

Dear list, I have some trouble generating a frequency table over a number of vectors. Creating these tables over simple numbers is no problem with table() > table(c(1,1,1,3,4,5)) 1 3 4 5 3 1 1 1 , but how can i for example turn: 0 1 0 0 0 1 0 1 0 1 0 0 0 1 0 1 0 0 into 0 0 1 1 1 0 0 2 0 1 0 3 My second problem is, sorting rows and columns of a matrix by the rowSums/colSums. I did it

"sum" (and perhaps other functions?) allows partial argument name matching after its three-dots argument: > sum(1:4, NA, n=78, na.rm=FALSE) [1] 10 > sum(1:4, NA, n=78, na.rm=TRUE) [1] 11 I can see there could be a discussion about whether or not this is a bug, but I think all will agree that it's a might peculiar. This is done in 2.0.1 but the same behavior is in 1.8.1.

#using "rollapply" to calculate a moving sum or running sum? #I am tryign to use rollapply to calcualte a moving sum? #I tried rollapply and get the error message #"Error in seq.default(start.at, NROW(data), by = by) : # wrong sign in 'by' argument" #example: mymatrix <- ( matrix(data=1:100, nrow=5, ncol=20) ) mymatrix_cumsum <- ( matrix(data=NA, nrow=5,

Hello List I have just written a little function that takes two matrices as arguments and returns a large matrix that is composed of the two input matrices in upper-left position and lower-right position with a padding value everywhere else. (function definition and toy example below). I need nonsquare matrices and rowname() and colname() inherited appropriately. Two questions: (1) Is there a

Hello all, Here goes one of my first functions. I want to make a nonparametric multiple sample comparison with unequal sample sizes (see Zar?s Biostatistical Analysis, 3rd. Ed., pg. 201 Example 10.11, pg. 288 Example 11.10). In the real world, I want to compare samples of fish length captured with different fishing gears. After using the Kruskal-Wallis test I want to check the differences

I am curious to know what algorithm R's mean function uses. Is there some reference to the numerical properties of this algorithm? I found the following C code in summary.c:do_summary(): case REALSXP: PROTECT(ans = allocVector(REALSXP, 1)); for (i = 0; i < n; i++) s += REAL(x)[i]; s /= n; if(R_FINITE((double)s)) { for (i = 0; i < n; i++) t += (REAL(x)[i] -

I'm observing that base::sum(x, na.rm=FALSE) for typeof(x) == "double" is much more time consuming when there are missing values versus when there are not. I'm observing this on both Window and Linux, but it's quite surprising to me. Currently, my main suspect is settings in on how R was built. The second suspect is my brain. I hope that someone can clarify the below