Online Program

Assessing the Validity of an Imputation Method Using Data from Comparable External Sources
Ellen Bouchery, Mathematica Policy Research, Inc. 
Eric Grau, Mathematica Policy Research, Inc. 
Julie Ingels, Mathematica Policy Research, Inc. 
*Sameena Maryam Salvucci, Mathematica Policy Research, Inc. 

Keywords: Item nonresponse, cold deck imputation, correlations

The Substance Abuse and Mental Health Services Administration (SAMHSA) designed and conducted the Survey of Revenues and Expenditures (SSR&E), a new survey with the purpose of collecting health expenditure information from a national sample of specialty substance abuse and mental health treatment facilities in the United States. Item response rates from sampled facilities in the SSR&E were lower than expected for some key items. These data were imputed using comparable data collected in two external universe surveys of these facilities that were conducted at a similar time. As a measure of the validity of the imputation method used, we will report on comparisons we made between responses by those facilities that reported values for these key items in both the sample survey and the external universe surveys.