NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED369815
Record Type: Non-Journal
Publication Date: 1994-Apr
Pages: 10
Abstractor: N/A
Reference Count: N/A
ISBN: N/A
ISSN: N/A
The Use of Control in Non-Randomized Designs.
Halperin, Si; Jorgensen, Randall
The concept of control is fundamental to comparative research. In research designs where randomization of observational units is not possible, control has been exercised statistically from a single covariate by a process of residualization. The alternative, known as subclassification on the propensity score, was developed primarily for biostatistical applications. The illustration included (comparing the physiological variables of 34 subjects from a home with a family history of hypertension with 46 subjects with no history of hypertension) demonstrates applicability to psychological research as well. Subclassification on the propensity score has several advantages over residualizing as a means of control. First, it allows control on many covariates. Second, it allows one to assess how well bias in the covariates has been reduced. Third, interactions with covariates can be followed by simple effects more readily than with residualization. Finally, subclassification on the propensity score requires fewer assumptions than residualization. Four tables. (Contains 19 references.) (Author/SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A