Accuracy of Obtaining Equivalent Groups by Matching Within Limits

Autor: Robert A. Rosemier
Rok vydání: 1967
Předmět:
Zdroj: The Journal of Experimental Education. 35:96-99
ISSN: 1940-0683
0022-0973
DOI: 10.1080/00220973.1967.11011005
Popis: For the past 40 years, references have been num erous (2, 5, 9) supportive of the concept of attaining pre-experimental equality of groups through random ization. Where such randomization is impossible or not feasible, alternatives have been promoted, such as "rotation experiments" (6,11), "blocking" (4) and analysis of covariance (2, 10, 12), rather than the traditional approach employing "matched groups". The realization that any matching which could be attained on one variable would not necessarily as sure matching on other related variables (if these could be identified), and R. L. Thorndike's discus sion (12) of the regression fallacies in the matched group experiment, would cause one to recognize the source of this favoritism. Hence, it would seem apparent that sufficient disclaimer of "matching" as an aid to valid inference has been served, and that more sound procedures have been offered as alter natives. Yet, even a superficial survey of non-tech nical journals in education and of on-going research planned by local school systems (and probably a great many master's and doctoral theses) would in dicate a stubborn, misleading, and continued r e 1 i ance on "matching" as the sole experimental design. With time, and increased persistence, perhaps educational research will benefit from the decay of this traditional approach in favor of the more recent innovations. Until then, any improvements in the presently popular techniques would seem to be worthwhile. The following is offered as one such at tempt. It is not intended as a long-term panacea for all of our experimental ills. Hopefully, it may be of some use to present neophytes, by providing a better insight into the practical problem s inherent in this design.
Databáze: OpenAIRE
Externí odkaz: