NonProbEst (version 0.2.4)

vd_weights: Calculates Valliant and Dever weights

Description

Computes weights from propensity estimates using the propensity stratification 1/p_i averaging formula introduced in Valliant and Dever (2011).

Usage

vd_weights(convenience_propensities, reference_propensities, g = 5)

Arguments

convenience_propensities

A vector with the propensities associated with the convenience sample.

reference_propensities

A vector with the propensities associated with the reference sample.

g

The number of strata to use; by default, its value is 5.

Value

A vector with the corresponding weights.

Details

The function takes the vector of propensities \(\pi(x)\) and calculates the weights to be applied in the Horvitz-Thompson estimator using the formula that can be found in Valliant and Dever (2019). The vector of propensities is divided in g strata (ideally five according to Cochran, 1968) aiming to have individuals with similar propensities in each strata. After the stratification, weight is calculated as follows for an individual i: $$w_i = \frac{n(g_i)}{ \sum_{k \in g_i} \pi_k (x)}$$ where \(g_i\) represents the strata to which i belongs, and \(n(g_i)\) is the number of individuals in the \(g_i\) strata.

References

Valliant, R., & Dever, J. A. (2011). Estimating propensity adjustments for volunteer web surveys. Sociological Methods & Research, 40(1), 105-137.

Cochran, W. G. (1968). The Effectiveness of Adjustment by Subclassification in Removing Bias in Observational Studies. Biometrics, 24(2), 295-313

Examples

# NOT RUN {
covariates = c("education_primaria", "education_secundaria")
data_propensities = propensities(sampleNP, sampleP, covariates)
vd_weights(data_propensities$convenience, data_propensities$reference)
# }