Compare Between Datat Sets Made Through Two Matching Merthods in Nonexperimental Data

• Activity Number: 350
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Gideon D. Bahn*+
• Companies: Loyola University Chicago/Hines VA Hospital
• Keywords: propensity matching method ; Fisher's matching method ; non-experimental study ; adjusted R square ; partial effect size ; compare matched datasets

Abstract

Causal inference is usually used in experimental studies with randomization. In medical research, randomization is often not feasible due to practical reasons. Non-randomized studies do not make causal conclusions between treatments and resulting outcomes, because there are potential confounding variables. Rubin (1974) and Rosenbaum and Rubin (1984) had started to draw causality in non-experimental studies, using matching method with propensity scores. The present study offers another matching method from Fisher's experimental design in a non-experimental data. In addition, the datasets made through both matching methods, propensity matching and Fisher's matching, are compared by evaluating adjusted R square and partial effect size. The results show that the matched dataset by Fisher's matching method provided both the adjusted R squares and partial effect size larger.