[I replied to Gabor only, I think i may be interesting to cc the list, for
the record.]
---------- Forwarded message ----------
From: bruno Piguet <bruno.pig...@gmail.com>
Date: 2010/6/29
Subject: Re: [R] Fast and simple tool for re-sampling of asynchronous time
series ?
To: Gabor Grothendieck <ggrothendi...@gmail.com>
2010/6/25 Gabor Grothendieck <ggrothendi...@gmail.com>
>
> The apply statements often have minimal performance advantage -- they
> are more for eliminating certain bookkeeping operations associated
> with loops and making the code more compact.
>
In this case, the real speed-up solution was a change in algorithm. If
time values are sorted (which is not a too unrealistic hypothesis), one
doesn't need to re-scan the whole vector to find the values close to the
next sample.
So, I eventually coded in R just as in C ou Fortran :
Y_sync <- rep(NA, length(Tx))
jmax <- length(Ty)
j0 <- 1
for (i in 1:length(Tx))
{
T_min <- Tx[i] - w
T_max <- Tx[i] + w
while ( j0<=jmax & Ty[j0] < T_min ) j0 = j0 + 1
if (j0 > jmax)
Y_sync[i] <- NA
else {
j1 <- j0
while ( j1<=jmax & Ty[j1] <= T_max) j1 = j1 + 1
Y_sync[i] <- mean(Y[j0:(j1-1)])
}
}
which gives something fast enough and simple enough for my current needs.
Bruno.
[[alternative HTML version deleted]]
______________________________________________
R-help@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.
