Open AccessResponse bias, weighting adjustments, and design effects in the Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS)
Author(s) -
Kessler Ronald C.,
Heeringa Steven G.,
Colpe Lisa J.,
Fullerton Carol S.,
Gebler Nancy,
Hwang Irving,
Naifeh James A.,
Nock Matthew K.,
Sampson Nancy A.,
Schoenbaum Michael,
Zaslavsky Alan M.,
Stein Murray B.,
Ursano Robert J.
Publication year - 2013
Publication title -
international journal of methods in psychiatric research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.275
H-Index - 73
eISSN - 1557-0657
pISSN - 1049-8931
DOI - 10.1002/mpr.1399
Subject(s) - population , psychology , sample (material) , weighting , psychological resilience , poison control , mental health , suicide prevention , demography , medicine , environmental health , psychiatry , social psychology , chemistry , chromatography , sociology , radiology
The Army Study to Assess Risk and Resilience in Servicemembers (Army STARRS) is a multi‐component epidemiological and neurobiological study designed to generate actionable recommendations to reduce US Army suicides and increase knowledge about determinants of suicidality. Three Army STARRS component studies are large‐scale surveys: one of new soldiers prior to beginning Basic Combat Training (BCT; n = 50,765 completed self‐administered questionnaires); another of other soldiers exclusive of those in BCT ( n = 35,372); and a third of three Brigade Combat Teams about to deploy to Afghanistan who are being followed multiple times after returning from deployment ( n = 9421). Although the response rates in these surveys are quite good (72.0–90.8%), questions can be raised about sample biases in estimating prevalence of mental disorders and suicidality, the main outcomes of the surveys based on evidence that people in the general population with mental disorders are under‐represented in community surveys. This paper presents the results of analyses designed to determine whether such bias exists in the Army STARRS surveys and, if so, to develop weights to correct for these biases. Data are also presented on sample inefficiencies introduced by weighting and sample clustering and on analyses of the trade‐off between bias and efficiency in weight trimming. Copyright © 2013 John Wiley & Sons, Ltd .