Skip to content

the standard error of the quantile

3 messages · Frank E Harrell Jr, swatch110362

Frank E Harrell Jr
# 
In small to moderate sample sizes, the Harrell-Davis quantile estimator is
more accurate than the ordinary sample quantile, and there is a good
standard error estimator for it using U-statistics.  See the hdquantile
function in the Hmisc package.
Frank
swatch110362 wrote:
-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/the-standard-error-of-the-quantile-tp4086479p4086494.html
Sent from the R help mailing list archive at Nabble.com.
swatch110362
# 
Thanks for your help.
But I need to estimator the standard error of the quantile in "survival
analysis", because my data is censored.
For example~


T<-c(84,240,261,332,348,437,521,565)
S<-c(0,1,1,0,1,0,1,0)  ##0 for censoed; 1 for event
G<-rep(1,8)

ori_s.surv<-survfit(Surv(T,S)~G)


--
View this message in context: http://r.789695.n4.nabble.com/the-standard-error-of-the-quantile-tp4086479p4086512.html
Sent from the R help mailing list archive at Nabble.com.
Frank E Harrell Jr
# 
Then see the latter part of my note.   I think also that, at least for the
median, the survival package will compute it more quickly for inclusion in a
bootstrap loop.  Note that you forgot to state require(survival) or
library(survival) below.
Frank
swatch110362 wrote:
-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/the-standard-error-of-the-quantile-tp4086479p4087199.html
Sent from the R help mailing list archive at Nabble.com.