On Sun, 23 Sep 2007, Rita Cristina Pinto Sousa wrote:
I?m using the package survey to obtain the statistics, fundamentally the
variance estimates. Can you explain why do I obtain the same result with the
replicate weights (as.svrepdesign function), for a stratified sample, and
without the replicate weights? I don?t understand it?
Without as.svrepdesign() you get the Horvitz-Thompson standard error
estimator for the total and linearization estimators for other statistics.
For the total and the mean these agree exactly with the JKn estimator.
For nonlinear statistics they do not agree exactly, although they are
usually close.
-thomas
Many thanks for your attention,
Rita Sousa.
***************************************
# Apuramentos IUTIC-E
rm(list=ls())
ano <- "2006"
require(survey)
options(survey.lonely.psu="remove")
setwd("C:/INE/Estat?stica-DME/IUTIC-E/")
#Read the DB
base_npc <-
read.table(paste(getwd(),"/Dados/",ano,"/NPC",".txt",sep=""),sep="\t",dec=",",header
= T)
base_npc[is.na(base_npc)]<-0
#Universe information
univ_npc <-
read.table(paste(getwd(),"/Dados/",ano,"/NPC_Univ",".txt",sep=""),sep="\t",dec=".",header
= T)
names(base_npc) <- tolower(names(base_npc))
names(univ_npc) <- tolower(names(univ_npc))
base_npc <- merge(base_npc,univ_npc,by="estr_cor")
#S? algumas vari?veis
#base_npc <-
base_npc[,c("npc","estr_cor","pond_npc","npc_univ","e_cuse","div1")]
#Survey design
desenho_npc <-
svydesign(id=~npc,strata=~estr_cor,weights=~pond_npc,fpc=~npc_univ,nest=T,data=base_npc)
#Replicate weights
desenho_npc_JK <- as.svrepdesign(desenho_npc,type="JKn")
#Without replicate weigths
svyby(~e_cuse,~div1,desenho_npc,svytotal,drop.empty.groups=FALSE,vartype=c("se","var","cvpct"))
#With replicate weigths
svyby(~e_cuse,~div1,desenho_npc_JK,svytotal,drop.empty.groups=FALSE,vartype=c("se","var","cvpct"))
******************************************
Citando Thomas Lumley <[EMAIL PROTECTED]>:
This message uses a character set that is not supported by the Internet
Service. To view the original message content, open the attached
message. If the text doesn't display correctly, save the attachment to
disk, and then open it using a viewer that can display the original
character set. <<message.txt>>
----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.
Thomas Lumley Assoc. Professor, Biostatistics
[EMAIL PROTECTED] University of Washington, Seattle
______________________________________________
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.
