Begin

Begin by loading the surveytable package. When you do, it will print a message explaining how to specify the survey that you’d like to analyze.

library(surveytable)

Now, specify the survey that you’d like to analyze.

set_survey(rccsu2018)

Check the survey name, survey design variables, and the number of observations to verify that it all looks correct.

For this example, we do want to turn on certain NCHS-specific options, such as identifying low-precision estimates. If you do not care about identifying low-precision estimates, you can skip this command. To turn on the NCHS-specific options:

set_opts(mode = "NCHS")
## * Mode: NCHS.

Alternatively, you can combine these two commands into a single command, like so:

set_survey(rccsu2018, mode = "NCHS")
## * Mode: NCHS.

Figure 1

This figure shows the percentage of residents by sex, race / ethnicity, and age group.

Sex.

tab("sex")

Race / ethnicity.

var_list("race")

tab("raceeth2")

In the published figure, the Hispanic and Other categories have been merged into a single category called “Another race or ethnicity”. We can do that using the var_collapse() function.

var_collapse("raceeth2"
             , "Another race or ethnicity"
             , c("Hispanic", "Other"))
tab("raceeth2")

Age group.

var_list("age")

age2 is a numeric variable. We need to create a categorical variable based on this numeric variable. This is done using the var_cut() function.

var_cut("Age", "age2"
        , c(-Inf, 64, 74, 84, Inf)
        , c("Under 65", "65-74", "75-84", "85 and over") )
tab("Age")

Figure 2

This figure shows the percentage of residents with Medicaid, overall and by age group.

tab("medicaid2")

As we can see, for some observations, the value of this variable is unknown (it’s missing or NA). The above command calculates percentages based on all observations, including the ones with missing (NA) values. However, in the published figure, the percentages are based on the knowns only. To exclude the NA’s from the calculation, use the drop_na argument:

tab("medicaid2", drop_na = TRUE)

Note that the table title alerts you to the fact that you are using known values only.

By age group:

tab_subset("medicaid2", "Age", drop_na = TRUE)

Note that according to the NCHS presentation criteria, some of the percentages should be suppressed.

Figure 4

(Figure 3 is slightly more involved, so we’ll do it next.)

• This figure shows the percentage of residents who have one of a select set of chronic conditions.
• In addition, it shows the distribution of residents by the number of conditions.

Here’s a table for high blood pressure.

tab("hbp")

Once again, unknown values (NA) are present, while the figure is based on knowns only. Therefore, we again will use the drop_na argument:

tab("hbp", "alz", "depress", "arth", "diabetes", "heartdise", "osteo"
    , "copd", "stroke", "cancer"
    , drop_na = TRUE)

class(rccsu2018$variables)
## [1] "data.frame"

rccsu2018$variables$num_cc = 0
for (vr in c("hbp", "alz", "depress", "arth", "diabetes", "heartdise", "osteo"
             , "copd", "stroke", "cancer")) {
  idx = which(rccsu2018$variables[,vr])
  rccsu2018$variables$num_cc[idx] = rccsu2018$variables$num_cc[idx] + 1
}
set_survey(rccsu2018, mode = "NCHS")
## * Mode: NCHS.

var_cut("Number of chronic conditions", "num_cc"
        , c(-Inf, 0, 1, 3, 10, Inf)
        , c("0", "1", "2-3", "4-10", "??"))
tab("Number of chronic conditions")

Figure 3

• This figure shows the percentage of residents who need help with one of the activities of daily living (ADLs).
• In addition, it shows the distribution of residents by the number of ADLs with which they need help.

Here’s a table for bathhlp (help with bathing):

tab("bathhlp")

This variable has multiple levels.

• Several of these levels correspond to a resident needing help,
• One level ("NEED NO ASSISTANCE") = does not need help
• One level ("MISSING") = unknown

We want to show (resident needing help) as a percentage of knowns only (that is, excluding the unknowns).

To do this, convert the variable to having 2 levels (needs help / does not need help) plus NA (for unknown); then use the drop_na argument to base percentages on knowns only.

for (vr in c("bathhlp", "walkhlp", "dreshlp", "transhlp", "toilhlp", "eathlp")) {
  var_collapse(vr
    , "Needs assistance"
    , c("NEED HELP OR SUPERVISION FROM ANOTHER PERSON"
      , "USE OF AN ASSISTIVE DEVICE"
      , "BOTH"))
  var_collapse(vr, NA, "MISSING")
}

tab("bathhlp", "walkhlp", "dreshlp", "transhlp", "toilhlp", "eathlp", drop_na = TRUE)

Now, go through the “advanced variable editing” steps – very similar to Figure 4 – to count how many ADLs were present.

rccsu2018$variables$num_adl = 0
for (vr in c("bathhlp", "walkhlp", "dreshlp", "transhlp", "toilhlp", "eathlp")) {
  idx = which(rccsu2018$variables[,vr] %in%
    c("NEED HELP OR SUPERVISION FROM ANOTHER PERSON"
      , "USE OF AN ASSISTIVE DEVICE"
      , "BOTH"))
  rccsu2018$variables$num_adl[idx] = rccsu2018$variables$num_adl[idx] + 1
}
set_survey(rccsu2018, mode = "NCHS")
## * Mode: NCHS.

For generating the figure, create a categorical variable based on num_adl, which is numeric.

var_cut("Number of ADLs", "num_adl"
        , c(-Inf, 0, 2, 6, Inf)
        , c("0", "1-2", "3-6", "??"))
tab("Number of ADLs")