Data Integration and Information Synthesis When Different Studies Have Similar but Not Exactly the Same Distribution

• Activity Number: 310 - Data Integration and Information Synthesis in Survival Analysis
• Type: Invited
• Date/Time: Tuesday, August 9, 2022 : 2:00 PM to 3:50 PM
• Sponsor: Lifetime Data Science Section
• Author(s): Jing Qin*
• Companies: Biostatistics Research Branch, National Institute of Allergy and Infectious Diseases
• Keywords: Biased results; Data integration; Heterogeneity; Selection bias

Abstract

The advance of technology makes data collection much easier than before. To assemble similar studies together for an enhanced inference for the underlying models and parameters has become indispensably important. It would be much easier to integrate information if all studies have the common distribution. However, in real life, this may not be the case. Similar studies may have something in common but may have subtle differences. For example, they may share common baseline hazard or density but have individual study specific parameters. To synthesize information, we have to make sure whether there is heterogeneity among different studies. If a discrepancy does exist, one has to take this into consideration when combing data together, otherwise, a biased result may occur. A few examples will be used for illustration.