Developing Generalized Variance Functions for Estimates of Recidivism Rates

• Activity Number: 445
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Karol Krotki* and Marcus Berzofsky and David Heller
• Companies: RTI International and RTI International and RTI International
• Keywords: Generalized Variance Function (GVF) ; Complex survey design ; Design effect ; Recidivism rates ; Online data tools

Abstract

This paper discusses the methods used to develop a generalized variance function (GVF) to properly compute standard errors (SEs) for recidivism rates for a cohort of inmates released from prison in 2005. GVFs use model-based parameters in a relatively simple function to calculate the SEs for an estimate while properly taking into account the complex survey design without requiring the use of statistical software. Thus, GVFs provide users a computationally easy method to calculate SEs for estimates from a survey with a complex design. The Bureau of Justice Statistics (BJS) has developed an online tool for estimating recidivism rates over time for a cohort of inmates released in 2005. The cohort of inmates was selected using a stratified design. Due to computational constraints the online tool could not use a statistical software package to dynamically compute requested SEs. Therefore, BJS wanted to incorporate a GVF into the tool to produce SEs along with the recidivism estimates. We discuss the models considered and the challenges associated with developing three parameter GVFs for each estimate type for this population.