Stratify: Randomly sub-sample data (stratified sampling)

In danielmarcelino/SciencesPo: A Tool Set for Analyzing Political Behavior Data

Description

The ability to randomly sub-sample population data "in silico" is essential for investigations such as sample effects on parameter estimation. This function sample row values of a data frame conditional to some strata attributes.

Usage

stratify(.data, group, size, select = NULL, replace = FALSE,
  both.sets = FALSE)

Arguments

.data	the data frame.
group	the grouping factor, may be a list.
size	the sample size.
select	if sampling from a specific group or list of groups.
replace	should sampling be with replacement?
both.sets	if TRUE, both 'sample' and '.data' are returned.

Examples

data(pollster2008)

# Let's take a 10% sample from all -PollTaker- groups in pollster2008
 stratify(pollster2008, "PollTaker", 0.1)

 # Let's take a 10% sample from only 'LV' and 'RV' groups from -Pop- in pollster2008
 stratify(pollster2008, "Pop", 0.1, select = list(Pop = c("LV", "RV")))

 # Let's take 3 samples from all -PollTaker- groups in pollster2008,
 # specified by column 1
stratify(pollster2008, group = 1, size = 3)

# Let's take a sample from all -Pop- groups in pollster2008, where we
# specify the number wanted from each group
stratify(pollster2008, "Pop", size = c(3, 5, 4))

# Use a two-column strata (-Pop- and -PollTaker-) but only interested in
# cases where -Pop- == 'LV'
stratify(pollster2008, c("Pop", "PollTaker"), 0.15, select = list(Pop = "LV"))

danielmarcelino/SciencesPo documentation built on Oct. 20, 2019, 1:15 a.m.