R help - Mar 2009 - CI from svyquantile in survey package

I am having trouble understanding (i.e. getting) confidence intervals from
   the survey package. I am using R version 2.8.1 (2008-12-22) and survey
   package (3.11-2) on FC7 linux. To simplify my question I use an example from
   that package:
   R> data(api)
   R> dclus1<-svydesign(id=~dnum, weights=~pw, data=apiclus1, fpc=~fpc)
   R>    (tst    <-    svyby(~api99,    ~stype,    dclus1,   svyquantile,
   quantiles=0.5,ci=TRUE))
     stype api99    se
   E     E   615 37.89
   H     H   593 69.52
   M     M   611 37.67
   R> str(tst)
   Classes ???svyby??? and 'data.frame':    3 obs. of  3 variables:
    $ stype: Factor w/ 3 levels "E","H","M": 1 2 3
    $ api99: num  615 593 611
    $ se   : num  37.9 69.5 37.7
    - attr(*, "svyby")=List of 7
     ..$ margins  : int 1
     ..$ nstats   : num 1
     ..$ vars     : int 1
     ..$ deffs    : logi FALSE
     ..$ statistic: chr "svyquantile"
     ..$ variables: chr "api99"
     ..$ vartype  : chr "se"
    -  attr(*,  "call")=  language  svyby.default(~api99, ~stype,
dclus1,
   svyquantile, quantiles = 0.5,      ci = TRUE)
   R>
   I do not see the CI in this structure.  Specifically, this problem came up
   when  I  was  working  through an example where after a similar set up
   information was extracted using
   R> tst$statistics.quantiles
   NULL
   R> tst$statistics.CIs
   NULL
   where of course the NULLs I get are due, in my case,  to the absences of the
   $statistics.XXXs.   The example did not display the structure so I could not
   compare my output with their output. (The example warns that these objects
   need to be unlisted which is not relevant here)
   Thanks in advance for any help
   Ron
-- 
R. R. Burns
Oceanside, CA

This is a consequence of changes made to support multiple imputation, where
extracting the standard errors was needed.  I will try to fix it. A work-around
is to use lapply() and subset()

lapply(c("E","M","H"), function(s)
svyquantile(~api99,subset(dclus1, stype==s), ci=TRUE,quantiles=0.5))

Thanks for reporting the problem.

       -thomas

On Wed, 11 Mar 2009, Ron Burns wrote:
>
>   I am having trouble understanding (i.e. getting) confidence intervals
from
>   the survey package. I am using R version 2.8.1 (2008-12-22) and survey
>   package (3.11-2) on FC7 linux. To simplify my question I use an example
from
>   that package:
>   R> data(api)
>   R> dclus1<-svydesign(id=~dnum, weights=~pw, data=apiclus1,
fpc=~fpc)
>   R>    (tst    <-    svyby(~api99,    ~stype,    dclus1,  
svyquantile,
>   quantiles=0.5,ci=TRUE))
>     stype api99    se
>   E     E   615 37.89
>   H     H   593 69.52
>   M     M   611 37.67
>   R> str(tst)
>   Classes ???svyby??? and 'data.frame':    3 obs. of  3 variables:
>    $ stype: Factor w/ 3 levels "E","H","M": 1
2 3
>    $ api99: num  615 593 611
>    $ se   : num  37.9 69.5 37.7
>    - attr(*, "svyby")=List of 7
>     ..$ margins  : int 1
>     ..$ nstats   : num 1
>     ..$ vars     : int 1
>     ..$ deffs    : logi FALSE
>     ..$ statistic: chr "svyquantile"
>     ..$ variables: chr "api99"
>     ..$ vartype  : chr "se"
>    -  attr(*,  "call")=  language  svyby.default(~api99, ~stype,
dclus1,
>   svyquantile, quantiles = 0.5,      ci = TRUE)
>   R>
>   I do not see the CI in this structure.  Specifically, this problem came
up
>   when  I  was  working  through an example where after a similar set up
>   information was extracted using
>   R> tst$statistics.quantiles
>   NULL
>   R> tst$statistics.CIs
>   NULL
>   where of course the NULLs I get are due, in my case,  to the absences of
the
>   $statistics.XXXs.   The example did not display the structure so I could
not
>   compare my output with their output. (The example warns that these
objects
>   need to be unlisted which is not relevant here)
>   Thanks in advance for any help
>   Ron
> --
> R. R. Burns
> Oceanside, CA
> ______________________________________________
> 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.
>
Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle