In order to provide national totals of graduate S&E students and postdoctoral appointees, data must be estimated for departments that provide partial responses and departments that do not respond at all.
Data imputation is an automated procedure that estimates data for nonrespondent and partially nonrespondent institutions and departments. Imputation involves calculating inflator/deflator factors for the key variables: total full-time students, total part-time students, total postdoctoral appointees, and total non-faculty doctoral researchers. The inflator/deflator factors reflect national growth/decline. The key variables come from fully responding departments in both the current and previous years' surveys for each combination of degree level and discipline. For a partially nonresponding department, after the key variables are calculated, they are distributed among the various subtotal and detail fields by using the same relative percentages last reported for that department. For a totally nonrespondent department, the inflator/deflator factors are applied to the previous year's key variable values for the department to derive a current year estimate.
Imputation of all data fields is performed for institutions that do not respond to the survey; partial imputation is performed for institutions omitting some of the questionnaire items. Since imputation is based on previously reported/imputed data, data for departments that are new to the survey and yet fail to provide any data cannot be imputed; these departments maintain a "nonrespondent" status in the survey database. Preliminary tables with data reported by S&E field are used to estimate data for any new departments that responds partially.
[ Survey Population | Data Collection And Processing Activities | Imputation Methodology ]