s-news
[Top] [All Lists]

bootstrapping a ratio

To: <s-news@lists.biostat.wustl.edu>
Subject: bootstrapping a ratio
From: "Manoel Pacheco" <mpacheco@glec-oh.com>
Date: Thu, 16 Dec 2004 12:45:18 -0500
Organization: GLEC
Reply-to: "Manoel Pacheco" <mpacheco@glec-oh.com>
I wrote the following code to generate confidence intervals for bootstraps
of ratios between two variables.

BAFp <- rlnorm(100, meanlog=0, sdlog=1)
BAFm <- rlnorm(100, meanlog=0, sdlog=1)
BAFdat <- data.frame(BAFp, BAFm)

# Bootstraps the distribution of ratios between two group means
ratios <- bootstrap(BAFdat, mean(BAFdat[,1])/mean(BAFdat[,2]))
ratios.ci <- limits.tilt(ratios)

plot(ratios)
qqnorm(ratios)
summary(ratios)

The script ran ok, but the output of plot(ratios) was unexpected. It was a
histogram of the frequency distribution of bootstrapped mean(BAFdat[,1])
values. How can I get a plot of the frequency distribution of bootstrapped
ratios?

As I am interested in examining changes in the confidence interval of
bootstrapped ratios along a series of increasing sample sizes, I attempted
to add a vector of block.sizes

ratios <- bootstrap(BAFdat, mean(BAFdat[,1]) / mean(BAFdat[,2]),
                              block.size = c(5, 10, 20, 40, 80))

All I got was a huge error message. Do I have to create a loop for such
analysis?

    Thanks,
                    Manolo


PS S+ Version 6.2, Windows 2000



<Prev in Thread] Current Thread [Next in Thread>