Creating replicate weights for a survey design object (with ANES 2016 data).

The following line loads data created with a previous R script or notebook.

load("anes-2016-prevote-desgn.RData")

The following makes use of the survey package. You may need to install it from CRAN using the code install.packages("survey") if you want to run this on your computer. (The package is already installed on the notebook container, however.)

library(survey)

Loading required package: grid

Loading required package: Matrix

Loading required package: survival


Attaching package: ‘survey’


The following object is masked from ‘package:graphics’:

    dotchart

The ‘automatic’ type gives jackknife replicates

anes_2016_prevote_jk <- as.svrepdesign(anes_2016_prevote_desgn,
                                       type="auto")

The number of replicates is determined by the number of clusters

anes_2016_prevote_jk

Call: as.svrepdesign(anes_2016_prevote_desgn, type = "auto")
Stratified cluster jackknife (JKn) with 65 replicates.

Here we select the multistage rescaled bootstrap

anes_2016_prevote_boo <- as.svrepdesign(anes_2016_prevote_desgn,
                                        type="mrbbootstrap")

Warning message in mrbweights(design$cluster, design$strata, design$fpc, ...):
“Design is sampled with replacement: only first stage used”

anes_2016_prevote_boo

Call: as.svrepdesign(anes_2016_prevote_desgn, type = "mrbbootstrap")
Multistage rescaled bootstrap with 50 replicates.

By default the number of bootstrap replicates is 50, we can change it to 200

anes_2016_prevote_boo <- as.svrepdesign(anes_2016_prevote_desgn,
                                        type="mrbbootstrap",
                                        replicates=200)

Warning message in mrbweights(design$cluster, design$strata, design$fpc, ...):
“Design is sampled with replacement: only first stage used”

anes_2016_prevote_boo

Call: as.svrepdesign(anes_2016_prevote_desgn, type = "mrbbootstrap", 
    replicates = 200)
Multistage rescaled bootstrap with 200 replicates.

A function to compute the percentage of 2012 Democratic and Republican voters who vote for a candidate of the same party in 2016:

StayerPerc <- function(weights,data){
    tab <- xtabs(weights~vote16+recall12,data=data)
    # Remove 'inap' responses
    tab <- tab[-6,-5]
    # Column percentages
    ptab <- 100*prop.table(tab,2)
    # The diagonal are the percentages of 'stayers'
    # among the voters of 2012.
    # The first two elements of the diagonal are
    # the Democratic and Republican stayers.
    structure(diag(ptab)[1:2],
              names=c("Democratic",
                      "Republican"))
}

Estimates and replication based standard errors based on jackknife

withReplicates(anes_2016_prevote_jk,
                StayerPerc)

            theta     SE
Democratic 68.765 3.8462
Republican 69.058 3.6052

withReplicates(anes_2016_prevote_boo,
                StayerPerc)

            theta     SE
Democratic 68.765 3.9823
Republican 69.058 3.3640