|Title||Impute: A SAS Application System for Missing Value Imputations-With Special Reference to HRS Income/Assets|
|Year of Publication||2001|
|Institution||Institute for Social Research, University of Michigan|
|City||Ann Arbor, Michigan|
|Keywords||Income, Methodology, Net Worth and Assets|
Survey questions that ask respondents to report amounts are often subject to high rates of missing data (Heeringa, Hill, and Howell, 1995; Hill, 1999; Juster and Smith, 1998). To avoid losing altogether valuable information from those who are unable or unwilling to provide answers to amount questions, more and more researchers rely on special questionnaire designs to collect bracket information, trying to infer exact amounts from the brackets collected. The work involving an inference of exact amount from bracket information may be called missing value imputation with brackets . When the bracket information is not available, it is reduced to a general type of imputation, called here missing value imputation without brackets . In this document, we describe a SAS application system, IMPUTE, designed for missing value imputation without brackets in general, and for inferring exact amount from bracket information in particular. To navigate our tour, and as a motivation for this project, we shall pay special attention to the Health and Retirement Study (HRS) income and assets imputations. Among many other advantages, the HRS income and asset data sets provide a good example for illustrating the uses of, and the limits to, the system.
|Endnote Keywords|| |
|Endnote ID|| |