An Investigation of the Behavior of Selected Variance Ratio Statistics Under Randomization for the Case of Disproportionate Subclass Numbers: A Thesis Submitted to the Faculty of the Graduate School

"It is frequently the case in educational and psychological experiments that, for a number of reasons, one is confronted with disproportionate numbers of observations in each of the treatment combinations of the experimental design. This imbalance may be a result of design specifications, practical limitations of the experimental situation or other conditions which randomly create inequality in frequency of treatment observations. There has been considerable debate as to which procedure is most appropriate under these circummstances. The parameters of interest in the present situation are taken to be independent of the numbers of observations in the various treatment combinations.It is the purpose of this investigation to study the behavior of the unweighted means variance-ratio statistic {Yates, 1934} under randomization. The randomization process does not require the usual normal theory assumptions and yields 'exact' results. In addition, the behavior of two least squares variance-ratio statistics based on the general linear model are studied using the randomization procedure." This book has 209 pages. The text contains NO internal marks whatsoever. Overall, this is a FINE copy of this MOST SCARCE title. Size: 4to - over 9¾" - 12" tall.…