R help - Oct 2009 - How do I vectorize this loop....

Basically I need to use the following data to calculate a squared error for
each Sample based on the expected Survival for the zone.

Basically, this code has Live/Dead for each sample, and I need to calculate
the square error based on the Expected Mean (ie, Survival).  The code looks
up the expectation for each zone and applies for each sample in the zone
using a loop:

Data1 <- data.frame(Sample=1:6, Live =c(24,25,30,31,22,23), Baseline
c(35,35,35,32,34,33),Zone =
c(rep("Cottonwood",3),rep("OregonAsh",3)))

Data2 <- data.frame(Zone = c("Cottonwood","OregonAsh"),
Survival c(0.83,0.76))

for (i in 1:nrow(Data1)) #####(Yi -Ybar*Yo)^2
	Data1$SquaredError[i] <- (Data1$Live[i] -
Data2$Survival[which(Data1$Zone[i]==Data2$Zone)]*Data1$Baseline[i])^2


My question is, can I vectorize this code to avoid the loop?  Obviously, I
could merge the 2 datasets first, but that would still require 2 steps and
Data1 would have a bunch of redundant data.  So, is there a better
alternative?  Is there some way I improve indexing syntax efficiency by
using rownames instead of a column vector?



-- 
View this message in context:
http://www.nabble.com/How-do-I-vectorize-this-loop....-tp26000933p26000933.html
Sent from the R help mailing list archive at Nabble.com.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com  
> -----Original Message-----
> From: r-help-bounces at r-project.org 
> [mailto:r-help-bounces at r-project.org] On Behalf Of chipmaney
> Sent: Wednesday, October 21, 2009 2:58 PM
> To: r-help at r-project.org
> Subject: [R] How do I vectorize this loop....
> 
> 
> Basically I need to use the following data to calculate a 
> squared error for
> each Sample based on the expected Survival for the zone.
> 
> Basically, this code has Live/Dead for each sample, and I 
> need to calculate
> the square error based on the Expected Mean (ie, Survival).  
> The code looks
> up the expectation for each zone and applies for each sample 
> in the zone
> using a loop:
> 
> Data1 <- data.frame(Sample=1:6, Live =c(24,25,30,31,22,23), Baseline
> c(35,35,35,32,34,33),Zone =
c(rep("Cottonwood",3),rep("OregonAsh",3)))
> 
> Data2 <- data.frame(Zone =
c("Cottonwood","OregonAsh"), Survival > c(0.83,0.76))
> 
> for (i in 1:nrow(Data1)) #####(Yi -Ybar*Yo)^2
> 	Data1$SquaredError[i] <- (Data1$Live[i] -
> Data2$Survival[which(Data1$Zone[i]==Data2$Zone)]*Data1$Baseline[i])^2
> 
> 
> My question is, can I vectorize this code to avoid the loop?
Does the following do what you want?
  > (Data1$Live -
Data2$Survival[match(Data1$Zone,Data2$Zone)]*Data1$Baseline)^2
  [1] 25.5025 16.4025  0.9025 44.6224 14.7456  4.3264
  
> Obviously, I
> could merge the 2 datasets first, but that would still 
> require 2 steps and
> Data1 would have a bunch of redundant data.
Why worry about a bunch of 'redundant' data?  Its space is
freed after it is no longer referenced, so if you do the
processing in a function and don't put all the temporarily
needed data in a permanent dataset it won't take up space
for long.  If you are running into the memory limits of your
hardware then you have to work harder to avoid using more
memory that you need.  (Note that vec[which(logical)] generally
gives the same result as vec[logical] but uses more space doing
so.)

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com 
 
> So, is there a better
> alternative?  Is there some way I improve indexing syntax 
> efficiency by
> using rownames instead of a column vector?
> 
> 
> 
> -- 
> View this message in context: 
> http://www.nabble.com/How-do-I-vectorize-this-loop....-tp26000
933p26000933.html
> Sent from the R help mailing list archive at Nabble.com.
> 
> ______________________________________________
> 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.
>