skip to main content skip to footer

Partial and Non-Response Bias Effects in a Nationwide Sample NSF

Creech, F Reid
Publication Year:
Report Number:
ETS Research Memorandum
Document Type:
Page Count:
Subject/Key Words:
National Science Foundation (NSF), Bias, Data Analysis, High School Seniors, Response Style (Tests), Sampling, Surveys


The non-response bias analysis of data from a stratified nationwide probability sample of high school seniors produced evidence in support of the hypothesis that nonrespondents tend to be of lower "educational level" than respondents. A partial-response bias analysis of the same data indicated that there were similarities between the biases of nonresponse and those of partial-response, but that different subsets of the partial response data were not necessarily consistent in the kind, amount, or direction of partial response bias. It was conjectured that similar inconsistencies might be found among nonrespondents, and that descriptors such as "educational level" might not be pervasive in characterizing differences between respondents and nonrespondents. Such terms as "educational level" probably require reification and objectivication as well; when educational level is defined as the number of semesters of coursework taken in certain designated subjects, higher numbers of courses taken are not always positively associated with the tendency to respond. It was also conjectured that there might be a continuum of completeness of responding, from nonrespondent through partial-respondent to full respondent with bias existing between any two points on the continuum.

Read More