qvar {gustave}

Quickly perform a variance estimation in common cases

Description

qvar (for "quick variance estimation") is a function performing analytical variance estimation in most common cases, that is:

• stratified simple random sampling

• non-response correction (if any) through reweighting

• calibration (if any)

Used with define = TRUE, it defines a so-called variance wrapper, that is a standalone ready-to-use function that can be applied to the survey dataset without having to specify the methodological characteristics of the survey (see define_variance_wrapper).

Usage

qvar(
  data,
  ...,
  by = NULL,
  where = NULL,
  alpha = 0.05,
  display = TRUE,
  id,
  dissemination_dummy,
  dissemination_weight,
  sampling_weight,
  strata = NULL,
  scope_dummy = NULL,
  nrc_weight = NULL,
  response_dummy = NULL,
  nrc_dummy = NULL,
  calibration_weight = NULL,
  calibration_dummy = NULL,
  calibration_var = NULL,
  define = FALSE,
  envir = parent.frame()
)

Arguments

data

The data.frame containing all the technical information required to prepare the variance estimation process (see other arguments below). Note that this file should contain all the units sampled, including the out-of-scope and non-responding units. If a variance estimation is to be performed right away (when define = FALSE), it should also contain the variables of interest.

...

One or more calls to a statistic wrapper (e.g. total(), mean(), ratio()). See examples and standard statistic wrappers

by

A qualitative variable whose levels are used to define domains on which the variance estimation is performed.

where

A logical vector indicating a domain on which the variance estimation is to be performed.

alpha

A numeric vector of length 1 indicating the threshold for confidence interval derivation (0.05 by default).

display

A logical verctor of length 1 indicating whether the result of the estimation should be displayed or not.

id

The identification variable of the units in data. It should be unique for each row in data and not contain any missing values.

dissemination_dummy

A character vector of length 1, the name of the logical variable in data indicating whether the unit does appear in the disseminated file and should be used for point estimates. It should not contain any missing values.

dissemination_weight

A character vector of length 1, the name of the numerical variable in data corresponding to the dissemination weight of the survey. It should not contain any missing values.

sampling_weight

A character vector of length 1, the name of the numeric variable in data corresponding to the sampling weights of the survey. It should not contain any missing values.

strata

A character vector of length 1, the name of the factor variable in data whose level match the stratification used in the survey. Character variables are coerced to factor. If defined, it should not contain any missing value. If NULL, the variance estimation process does not take any stratification into account.

scope_dummy

A character vector of length 1, the name of the logical variable in data indicating whether the unit belongs to the scope of the survey or not. Numerical variables are coerced to logical. If defined, it should not contain any missing value. If NULL, all units are supposed to be within the scope of the survey.

nrc_weight

A character vector of length 1, the name of the numerical variable in data corresponding to the weights after non-response correction. If defined, all responding units should have a non-missing value. If NULL, all units are supposed to be responding and the variance estimation process does not include a second phase in order to take non-response into account.

response_dummy

A character vector of length 1, the name of of the logical variable in data indicating whether the unit is a responding unit or not. Numerical variables are coerced to logical. response_dummy should be defined as long as a nrc_weight is provided. All units in the scope of the survey should have a non-missing value.

nrc_dummy

A character vector of length 1, the name of the logical variable in data indicating whether the units did take part in the non-response correction process. All units in the scope of the survey should have a non-missing value.

calibration_weight

A character vector of length 1, the name of the numerical variable in data corresponding to the calibrated weights. If defined, all responding units should have a non-missing value. If NULL, the variance estimation process does not take any calibration step into account.

calibration_dummy

A character vector of length 1, the name of of the logical variable in data indicating whether the unit did take part in the calibration process or not. Numerical variables are coerced to logical. If defined, all responding units should have a non-missing value. If NULL, calibration is supposed to have been performed on all responding units.

calibration_var

A character vector, the name of the variable(s) used in the calibration process. Logical variables are coerced to numeric. Character and factor variables are automatically discretized. calibration_var should be defined as long as a calibration_weight is provided. All units taking part in the calibration process should have only non-missing values for all variables in calibration_var.

define

Logical vector of lentgh 1. Should a variance wrapper be defined instead of performing a variance estimation (see details and examples)?

envir

An environment containing a binding to data.

Details

qvar performs not only technical but also methodological checks in order to ensure that the standard variance estimation methodology does apply (e.g. equal probability of inclusion within strata, number of units per stratum).

Used with define = TRUE, the function returns a variance estimation wrapper, that is a ready-to-use function that implements the described variance estimation methodology and contains all necessary data to do so (see examples).

Note: To some extent, qvar is analogous to the qplot function in the ggplot2 package, as it is an easier-to-use function for common cases. More complex cases are to be handled by using the core functions of the gustave package, e.g. define_variance_wrapper.

See Also

define_variance_wrapper, standard_statistic_wrapper

Examples

### Example from the Information and communication technologies (ICT) survey

# The (simulated) Information and communication technologies (ICT) survey 
# has the following characteristics: 
# - stratified one-stage sampling design
# - non-response correction through reweighting in homogeneous response groups
# - calibration on margins.

# The ict_survey data.frame is a (simulated) subset of the ICT 
# survey file containing the variables of interest for the 612
# responding firms.

# The ict_sample data.frame is the (simulated) sample of 650
# firms corresponding to the ict_survey file. It contains all
# technical information necessary to estimate a variance with
# the qvar() function.

## Methodological description of the survey

# Direct call of qvar()
qvar(

  # Sample file
  data = ict_sample,
  
  # Dissemination and identification information
  dissemination_dummy = "dissemination",
  dissemination_weight = "w_calib",
  id = "firm_id",
  
  # Scope
  scope_dummy = "scope",
  
  # Sampling design
  sampling_weight = "w_sample", 
  strata = "strata",
  
  # Non-response correction
  nrc_weight = "w_nrc", 
  response_dummy = "resp", 
  hrg = "hrg",
  
  # Calibration
  calibration_weight = "w_calib",
  calibration_var = c(paste0("N_", 58:63), paste0("turnover_", 58:63)),
  
  # Statistic(s) and variable(s) of interest
  mean(employees)
 
)

# Definition of a variance estimation wrapper
precision_ict <- qvar(

  # As before
  data = ict_sample,
  dissemination_dummy = "dissemination",
  dissemination_weight = "w_calib",
  id = "firm_id",
  scope_dummy = "scope",
  sampling_weight = "w_sample", 
  strata = "strata",
  nrc_weight = "w_nrc", 
  response_dummy = "resp", 
  hrg = "hrg",
  calibration_weight = "w_calib",
  calibration_var = c(paste0("N_", 58:63), paste0("turnover_", 58:63)),
  
  # Replacing the variables of interest by define = TRUE
  define = TRUE
  
)

# Use of the variance estimation wrapper
precision_ict(ict_sample, mean(employees))

# The variance estimation wrapper can also be used on the survey file
precision_ict(ict_survey, mean(speed_quanti))

## Features of the variance estimation wrapper

# Several statistics in one call (with optional labels)
precision_ict(ict_survey, 
  "Mean internet speed in Mbps" = mean(speed_quanti), 
  "Turnover per employee" = ratio(turnover, employees)
)

# Domain estimation with where and by arguments
precision_ict(ict_survey, 
  mean(speed_quanti), 
  where = employees >= 50
)
precision_ict(ict_survey, 
  mean(speed_quanti), 
  by = division
)

# Domain may differ from one estimator to another
precision_ict(ict_survey, 
  "Mean turnover, firms with 50 employees or more" = mean(turnover, where = employees >= 50),
  "Mean turnover, firms with 100 employees or more" = mean(turnover, where = employees >= 100)
)

# On-the-fly evaluation (e.g. discretization)
precision_ict(ict_survey, mean(speed_quanti > 100))

# Automatic discretization for qualitative (character or factor) variables
precision_ict(ict_survey, mean(speed_quali))

# Standard evaluation capabilities
variables_of_interest <- c("speed_quanti", "speed_quali")
precision_ict(ict_survey, mean(variables_of_interest))

# Integration with %>% and dplyr
library(magrittr)
library(dplyr)
ict_survey %>% 
  precision_ict("Internet speed above 100 Mbps" = mean(speed_quanti > 100)) %>% 
  select(label, est, lower, upper)