New subject: [gov2001-l] writing function for A

12 Feb 2007

Hi,

I?m not sure whether I?m missing something ?obvious? (retrospectively, of
course) with the sample() function:

What I?ve got for d) so far is:

beta.ests  <- matrix (ncol=4, nrow=100) # 1st col == 100 beta.zero
estimates, 2nd col == 100 beta.1 estimates, and so on...

colnames(beta.ests) <- c("beta.0", "beta.1", "beta.2",
"beta.3")

N <- nrow(data) # number of observations

for (i in 1:100){

  # sample can't be used directly (I think?) because it returns an element
from a vector

  # rather than a row from a dataframe or matrix. However, the elements
returned can then be used

  # to select the rows to use in a sampled subset

  sample.selector <- sample(1:N, 50, replace=T) # select 50 rows between [1,
1000] allowing the same row to be selected more than once

  sampled.subset <- data[sample.selector,] # selects rows

  sample.lm <- lm(y ~ x1 + x2 + x3, data="sampled.subset") #constructs
linear model based on sample

  beta.ests[i,] <- as.numeric(sample.lm$coefficients) #extracts the betas
from the lm object and places in rows on beta.ests

}

rm(i)

The problem, I?ve discovered is with the following:

  sampled.subset <- data[sample.selector,]

because sample.selector could contain the same number twice and the rows to
select are only meant to be specified once. 

I know this *could* be sorted (I think) by stripping sample.selector to only
include unique values, then appending nonunique rows to the sample dataframe
the required number of times afterwards, but is there a more appropriate
(e.g. one/two command) way of doing this instead?

Thanks in advance for your responses...

Jon Minton

-- 
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.5.441 / Virus Database: 268.17.36/681 - Release Date: 11/02/2007
18:50

[gov2001-l] using sample()