CODEWITHSUNDEEP
rdataframesurvey
vrad
vrad

Reputation: 53

Confidence Intervals for Weighted Survey Analysis

I'm interested in using confidence intervals calculated from the survey package, as is done here:

data(api)
dclus1<-svydesign(id=~dnum, fpc=~fpc, data=apiclus1)

df <- svyciprop(~I(ell==0), dclus1, method="li")
str(df)

I am having trouble assigning this to an object for later use, though, and it seems as if the only way I can get it out is by copying and pasting, which can't possibly be true.

How do I create a dataframe from this object so that I can use the confidence intervals elsewhere?

I would like an output table that looks like this:

Variable Value Estimate Low_ci Hi_ci
ell 0 0.02185 0.000664 0.11
ell 1 0.0273 0.0034 0.09

Upvotes: 2

Views: 374

Answers (1)

Thomas Lumley
Thomas Lumley

Reputation: 2765

You can extract the point estimate with as.numeric and the confidence interval with confint

> data.frame(p=as.numeric(df), ci=confint(df))
                     p      ci.2.5.  ci.97.5.
I(ell == 0) 0.02185792 0.0006639212 0.1077784

I don't know what you mean about a data frame with a row for ell==1: ell is not a binary variable: it's a percentage, with quite a lot of zeros

> summary(apiclus1$ell)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   0.00   16.00   26.00   27.61   37.00   73.00

If you had a set of confidence intervals to compute, you might use svyby to get them in a data frame. For example, the proportion of ell==0 by school type:

> svyby(~I(ell==0),~stype, design=dclus1, svyciprop,vartype="ci")
  stype I(ell == 0)        ci_l      ci_u
E     E  0.01388889 0.001383745 0.1252329
H     H  0.07142857 0.010052692 0.3681672
M     M  0.04000000 0.003850809 0.3099199
> is.data.frame(svyby(~I(ell==0),~stype, design=dclus1, svyciprop,vartype="ci"))
[1] TRUE

Upvotes: 1

Related Questions