RDU Webinar Series: Use of Imputation Techniques for Modeling Missing Data in Population-based Cancer Registries

Recorded On: 05/20/2016

The purpose of this webinar is to give a global overview of missing data issues and several different imputation techniques. We will also illustrate various applications of imputation techniques in population-based cancer registry settings. We will show improving completeness of biomarkers [e.g., estrogen receptor (ER) status] for breast cancer cases from population-based cancer registries through modeling techniques using standard assumption such as ER status is missing at random (MAR). Additionally, we will demonstrate use of novel imputation methods to perform sensitivity analyses if one suspects ER status is missing not at random (MNAR). The imputed databases are available to researchers for conducting a variety of analyses of breast cancer incidence trends through SEER*Stat software.

Topic 1: Overview of Missing Data Issues and Several Different Imputation Techniques.

Topic 2: Imputing Missing Estrogen Status (ER) Status for Breast Cancer Cases from Population-based Cancer Registries

Topic 3: Imputing Estrogen Receptor (ER) status under a Missing Not At Random (MNAR) Assumption

Presenters:

Barnali Das, PhD
Senior Statistician
Westat

Nadia Howlader, MS
Mathematical Statistician
Surveillance Research Program, National Cancer Institute

Rebecca Andridge, PhD
Assistant Professor, Division of Biostatistics
Ohio State University College of Public Health

Key:

Complete
Failed
Available
Locked
RDU Webinar Series: Use of Imputation Techniques for Modeling Missing Data in Population-based Cancer Registries