Title: Creating Sampling Weights through an Iterative Raking Procedure
Authors: Sevigny, Mitch - Craig Hospital ( Presenting ) Ketchum, Jessica - Craig Hospital
Abstract: Many data centers rely on using survey sampling for their data collection. This information is meant to be representative of a specific target population. However issues can arise like non-responsive subjects and sampling fluctuations where this doesn’t occur. One way of accounting for this is to develop sampling weights that allow the sample to better resemble the target population. An example of this involves the Traumatic Brain Injury (TBI) Model Systems National Database (NDB). The TBI Model Systems’ population has multiple demographic and injury characteristics that are not consistent with the national TBI population. Using an iterative raking procedure, weights have been created at the TBI Model Systems NDB to adjust for these differences. This presentation will discuss how this iterative raking process is conducted and how it can be utilized for other databases. Guidance will also be included on using methods like weight trimming and adjusting for non-response to improve population-based inference.
|