null
Stanford School of Medicine
Model Site Optional secondary text for expansion
on name or identity only

Quantitative Sciences Unit Research Methods Seminar - Multiple Imputation in Practice -- Approaches for handling categorical and interaction variables

May 07, 2013 (Tue) | 4:00 PM -5:00 PM
1070 Arastradero #109 : Palo Alto, CA

Missing data is a pervasive problem in medical and epidemiological research. Multiple imputation (MI), a simulation-based method, is one reasonable approach for handling missing data. Recently, mainstream statistical packages such as SAS, STATA and R have incorporated MI procedures allowing easy access to its use. While in theory MI yields valid results when data are missing at random (MAR), in practice the story is more nuanced. Much of the burden remains on the user for appropriate application in order for validity to hold. For example, MI is not completely straightforward in the presence of categorical variables, derived variables or interaction terms. Further, different software packages not only rely on different methods for imputation but also make different options available for han

Department:  Medicine

Contact: Linda Enomoto | 650-723-5096 | linda.enomoto@stanford.edu

Presenter(s):

  • Aya Mitani QSU Biostatistician

        Course Website

Stanford Medicine Resources:

Footer Links: