PS-matching.qmd

---
title: "Code example: ATE and HTE"
author: 'Marlon, Robin und Joel'
date: "2024-07-07"
output: html_document
---

```{r}
#| warning: false
library(MatchIt)
```

```{r}
# Simulierte Daten
set.seed(123)
n <- 1000
age <- rnorm(n, 30, 10)
education <- rnorm(n, 12, 2)
income <- 5000 + 500 * age + 1000 * education + rnorm(n, 0, 1000)

# Treatment-Indikator (Teilnahme am Programm)
treatment <- rbinom(n, 1, plogis(0.3 * age - 0.5 * education))

# Einkommen nach Behandlung
income_post <- income + treatment * (500 + (100 * age)) + rnorm(n, 0, 1000)

# Dataframe erstellen
data <- data.frame(age, education, treatment, income_post)

# 1. Matching anhand der Kovariaten durchführen, wobei der Propensity Score intern berechnet wird
match_it <- matchit(treatment ~ age + education, method = "nearest", data = data, caliper = 0.1)

# 2. Gematchte Daten extrahieren
matched_data <- match.data(match_it)

# 3. Überprüfen der Balance der Kovariaten
print(summary(match_it))

# 4. Visualisierung der Balance der Kovariaten nach Matching
library(ggplot2)
ggplot(matched_data, aes(x = factor(treatment), y = age)) +
  geom_boxplot() +
  labs(title = "Verteilung des Alters nach Treatment", x = "Treatment", y = "Age")

ggplot(matched_data, aes(x = factor(treatment), y = education)) +
  geom_boxplot() +
  labs(title = "Verteilung der Bildung nach Treatment", x = "Treatment", y = "Education")

# 5. Schätzen des Behandlungseffekts
ate_model <- lm(income_post ~ treatment, data = matched_data)
summary(ate_model)

#print(match_it)
```