5 messages · r student, David Winsemius, Thomas Lumley
I'm trying to create a density plot using census data, where the weights don't sum to 1.
plot(density(oh$FINCP,weights=oh$PWGTP))
Warning message: In density.default(oh$FINCP, weights = oh$PWGTP) : sum(weights) != 1 -- will not get true density How would I go about doing this? Thanks!
I'm trying to create a density plot using census data, where the weights don't sum to 1.
plot(density(oh$FINCP,weights=oh$PWGTP))
Warning message: In density.default(oh$FINCP, weights = oh$PWGTP) : sum(weights) != 1 -- will not get true density How would I go about doing this?
Wouldn't you just divide by the sum?
David Winsemius, MD West Hartford, CT
Like below? plot(density(oh$FINCP,weights=oh$PWGTP/sum(oh$PWGTP)))
On Aug 2, 2011, at 12:51 PM, r student wrote:
I'm trying to create a density plot using census data, where the weights don't sum to 1.
plot(density(oh$FINCP,weights=oh$PWGTP))
Warning message: In density.default(oh$FINCP, weights = oh$PWGTP) : ?sum(weights) != 1 ?-- will not get true density How would I go about doing this?
Wouldn't you just divide by the sum? -- David Winsemius, MD West Hartford, CT
Like below? plot(density(oh$FINCP,weights=oh$PWGTP/sum(oh$PWGTP)))
I don't understand why you are asking for approval. You are the one with the data and know where they came from. We have none of that background.
David. > On Tue, Aug 2, 2011 at 10:06 AM, David Winsemius <dwinsemius at comcast.net > > wrote: >> >> On Aug 2, 2011, at 12:51 PM, r student wrote: >> >>> I'm trying to create a density plot using census data, where the >>> weights don't sum to 1. >>> >>> >>>> plot(density(oh$FINCP,weights=oh$PWGTP)) >>> >>> >>> Warning message: >>> In density.default(oh$FINCP, weights = oh$PWGTP) : >>> sum(weights) != 1 -- will not get true density >>> >>> >>> How would I go about doing this? >> >> Wouldn't you just divide by the sum? >> >> -- >> >> David Winsemius, MD >> West Hartford, CT >> >> David Winsemius, MD West Hartford, CT
Like below? plot(density(oh$FINCP,weights=oh$PWGTP/sum(oh$PWGTP)))
Yes
If you are doing lots of analyses with weighted data you might want to
look at the survey package. It also has a density estimator, in
svysmooth(), which works very much the same way as density() for
weighted data, but doesn't complain about rescaling the weights.
-thomas
Thomas Lumley Professor of Biostatistics University of Auckland