stein_beta: Calculate Stein's Beta for Each Cluster

In DHSr: Create Large Scale Repeated Regression Summary Statistics Dataset and Visualization Seamlessly

Calculate Stein's Beta for Each Cluster

Description

This function calculates Stein's Beta for each cluster within the dataset. It applies Stein's shrinkage estimator to the specified beta estimates within each cluster.

Usage

stein_beta(data, cluster_id, beta)

Arguments

data	A dataframe containing the data.
cluster_id	The name of the column representing the cluster IDs.
beta	The name of the column representing the beta estimates.

Value

A data frame containing the input data with additional columns:

stein_beta	The Stein-shrinkage adjusted beta values.
lambda_d	Shrinkage factors for each cluster.
mu_beta_m	Mean beta values for each cluster.
sigma_hat_sq	Estimated variance of the beta values within clusters.
sum_of_squares	Sum of squared deviations of beta values from their mean.

Examples

# Create dummy data
library(dplyr)
set.seed(123)
dummy_data <- data.frame(
  years_education = rnorm(100, 12, 3),    # Represents years of education
  gender_female = rbinom(100, 1, 0.5),    # 1 = Female, 0 = Male
  household_wealth = sample(1:5, 100, replace = TRUE),  # Wealth index from 1 to 5
  district_code = sample(1:10, 100, replace = TRUE)     # Represents district codes
) %>% arrange(district_code)

# Define a regression formula
formula <- years_education ~ gender_female + household_wealth + household_wealth:gender_female

# Run the regression for all districts
results1 <- DHSr::Replm2(dummy_data, formula, "district_code", "normal")

# Assign random clusters for demonstration
clusters <- data.frame(
  district_code = unique(dummy_data$district_code),
  cluster_id = sample(1:3, length(unique(dummy_data$district_code)), replace = TRUE)
)

# Merge clusters with regression results
cluster_beta <- merge(clusters, results1, by.x = "district_code", by.y = "location")

# Apply Stein Beta shrinkage
results_with_stein_beta <- DHSr::stein_beta(
  data = cluster_beta,
  cluster_id = "cluster_id",                # Column for cluster IDs
  beta = "estimate_gender_female"          # Column for beta estimates
)

# View results
print(head(results_with_stein_beta))

DHSr documentation built on April 4, 2025, 12:18 a.m.