Usage
nsum.internal.validation(survey.data, known.popns = NULL, total.popn.size = NULL, degrees = NULL, missing = "ignore", kp.method = FALSE, weights = NULL, killworth.se = FALSE, return.plot = FALSE, verbose = FALSE, bootstrap = FALSE, ...)
Arguments
survey.data
the dataframe with the survey results
known.popns
if not NULL, a vector whose entries are the size of the known
populations, and whose names are the variable names in the dataset
corresponding to each one. if NULL, then assume that the survey.data
dataframe has an attribute called 'known.popns' containing this vector.
total.popn.size
the size of the entire population. if NA,
this function works with proportions;
if NULL, it looks for the 'total.popn.size' attribute of the
dataset survey.data;
if not NULL or NA, it
works with absolute numbers (ie, the proportions * total popn size)
degrees
if not NULL, then the name or index of the column in the datset
containing the degree estimates. if NULL, then use the known population
method to estimate the degrees (see kp.degree.estimator) missing
if "ignore", then proceed with the analysis without
doing anything about missing values. if "complete.obs"
then only use rows that have no missingness for the
computations (listwise deletion). care
must be taken in using this second option
kp.method
if TRUE, then we're using known population method estimates of the
degrees. this means we have to recompute the degrees each time we
hold out a known subgroup. if the degrees come from another estimator,
like the summation method, then we don't need to do that since we
don't use the ARD questions in coming up with the degree estimate.
weights
if not NULL, weights to use in computing the estimate. this
should be the name of the column in the survey.data which has
the variable with the appropriate weights. these weights
should be construted so that, eg, the mean of the degrees is
estimated as (1/n) * \sum_i w_i * d_i
killworth.se
if TRUE, return the Killworth et al estimate of the standard error
return.plot
if TRUE, make and return a ggplot2 plot object
verbose
if TRUE, report more detailed information about what's going on
bootstrap
if TRUE, use bootstrap.estimates to take bootstrap resamples
in order to obtain intervals around each estimate. in this case,
you are expected to also pass in at least bootstrap.fn,
survey.design, and num.reps
...
additional arguments, which are passed on to bootstrap.estimates
if bootstrap is TRUE