Anthony is right about the cause. The lack of partial=TRUE option in
rake is deliberate, and there's really only such an option in
postStratify() for debugging or simulation purposes. What
partial=TRUE allows is the situation where an observation in your
sample has a combination of values that doe
(1) you hit a memory error, because you are including too many variable
levels. even if you narrowed your 16 variables down to these four, the
`rake` function would need room for matricies containing as much data as a
6-million record table:
nrow( expand.grid( data473t[ , c( 'm11' , 'm12c' , 'm13
Here is some code with a subset of 50 cases that produces a similar set
of errors to what I got with the full data set (31,690 cases):
##
# Load "survey" package
#
library(survey)
##
# Create raking margins
##
pop.m01 <- data.frame(m01=1:14,
Freq=c(1013620, 136
could you provide a reproducible example ?dput is your friend
http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example
On Sat, Jun 7, 2014 at 11:22 AM, Michael Willmorth <
mwillmo...@clearwater-research.com> wrote:
> I'm teaching myself how to use rake() in the R "
I'm teaching myself how to use rake() in the R "survey" package, using Thomas
Lumley's "Complex Surveys" book. Working with one of my data sets, I received
an error that has me stumped.
Here is the code I used:
d473.raked <- rake(sdes473,
sample=list(~m01, ~m02, ~m03, ~m04, ~m05, ~m06c, ~m07c,
5 matches
Mail list logo
