povmap (version 1.0.1)

ebp_compute_cv: Coefficient of Variation (CV) estimations for Unit EBP Model Headcount Estimates

Description

Function ebp_compute_cv estimates CVs for the headcount of the unit model EBP functions using three different methods. CV, by definition, is the ratio of mean square error of the head count to the head count estimates. Therefore, the CV types are distinguished by the method of estimating the mean square.

Usage

ebp_compute_cv(
  model,
  calibvar = NULL,
  boot_type = "calibrate",
  designvar = NULL,
  threshold = NULL,
  B = model$call$B
)

Value

dataframe containing different types of CV values for the headcount

Arguments

model

an object returned by the ebp function of type "emdi ebp", representing point and MSE estimates

calibvar

the calibration variable to be used in method 1

boot_type

the bootstrap type "calibrated" or "naive" to be used in method 1

designvar

the survey design variable to be used in estimating the design effect for method 3.

threshold

a number defining a threshold. The argument defaults to NULL. In this case, the threshold is set to 60% of the median of the variable that is selected as dependent variable similary to the at-risk-of-poverty rate used in the EU (see also Social Protection Committee 2001). However, any desired threshold can be chosen.

B

number of bootstrap iterations for variance estimation. Defaults to number of bootstrap iteration in ebp obeject (specified in model).

Details

Method 1 uses the calibrated/naive bootstrapping of the MSE which allows to calibrate each bootstrap sample on auxiliary information using the direct function.' Calibrated bootstrap improves on the bias of the naive bootstrap when used in the complex survey context (see Rao and Wu (1988)) for more details.

Method 2 employs the Horowitz Thompson variance estimation technique to compute MSE i.e. each household is assigned the probability selection within the sample under a given sampling scheme. The computation employs sae::direct function.

Method 3 finally uses the design effect adjusted naive calibrated MSE. The design effect is estimated using the survey::svydesign function.

Examples

# \donttest{
data("eusilcA_pop")
data("eusilcA_smp")

# estimate a unit model
ebp_model <- ebp(fixed = eqIncome ~ gender + eqsize + cash +
                    self_empl + unempl_ben + age_ben + surv_ben + sick_ben +
                    dis_ben + rent + fam_allow + house_allow + cap_inv +
                    tax_adj,
                 pop_data = eusilcA_pop, pop_domains = "district",
                 smp_data = eusilcA_smp, smp_domains = "district",
                 na.rm = TRUE, weights = "weight",
                 pop_weights = "hhsize", MSE = TRUE, weights_type = "nlme",
                 B = 2, L = 2)

# compute CV table
ebp_compute_cv(model = ebp_model, calibvar = "gender")
# }