Propensity Score Matching, Adjustment, and Randomized Experiments

This course focuses on testing whether the results of a program can be attributed to a given cause. For example, was the increase in customer sales due to mailing of sales flyers? Was the health improvement due to the new medication? What conclusion can be drawn? The following cases are examined: randomized controlled experiments and observational studies that require adjustment to reduce bias by using propensity score analysis through either propensity score matching or propensity score adjustment. You will learn how to identify situations in which the simple method of multiple linear regression is inadequate and how to apply quasi-experimental analysis methods to real-world data for the following techniques: Propensity Score Matching and Propensity Score Adjustment.

Voraussetzungen
Before attending this course, you should complete or have the equivalent working experience of the following courses: „SAS Programming 1: Grundlagen" (PRG1) and „Statistik 1: Varianzanalyse, Regression und logistische Regression" (ST193).

Zielgruppe
Data analysts or statisticians in the fields of finance, telecommunications, pharmaceuticals, retail, and the public sector, who have an understanding of basic statistics and SAS programming.

Module
Base SAS Software

Kursinhalte

• Introduction to Causation
  • Program evaluation
  • Evaluation designs (optional)
  • Checklist for evaluations (optional)
• Randomized Experiments
  • Multiple linear regression
  • Issues with randomized design
  • Endogeneity
• Propensity Score Matching: Theory and Practice
  • Real-world non-random treatment assignment
  • Interpreting the propensity matched results
• Propensity Score Adjustment: Theory and Practice
  • Examples of real-world propensity score adjustment
  • Motivation behind propensity score adjustment
  • Real-world practice using banking example
• Group Discussion of Real-World Examples
  • Discuss examples of real-world quasi-experimental designs
  • Real-world data from students

Referent
Howard S. Friedman, Ph.D., professor Columbia University and partner DataMed Solutions LLC, or Paul Thurman, professor Columbia University
Dieser Kurs ist Bestandteil folgender Rolle(n):

• Market Researcher

zurück...