Open AccessSOME APPLICATIONS OF PROCEDURES FOR ALLOCATING TESTING TIME 1
Author(s) -
Novick Melvin R.,
Thayer Dorothy T.
Publication year - 1969
Publication title -
ets research bulletin series
Language(s) - English
Resource type - Journals
eISSN - 2333-8504
pISSN - 0424-6144
DOI - 10.1002/j.2333-8504.1969.tb00161.x
Subject(s) - computer science , unit (ring theory) , regression , homogeneous , computation , regression analysis , optimal allocation , mathematical optimization , statistics , mathematics , algorithm , machine learning , combinatorics , mathematics education
Detailed computations and data analysis using the algorithm developed by Woodbury and Novick (1968) and the procedure developed by Jackson and Novick (1969) for allocating testing time, when using regression and unit weights respectively, are presented, using data especially selected to illustrate how these techniques overcome the major difficulties, both computational and methodological, that may arise in such analyses. Several further sets of data are analyzed to compare the composite scale correlations obtained with a criterion using combinations of original or optimal time allocations and unit or optimal weights for homogeneous subscales. The primary purpose of this paper is methodological rather than substantive, that is, to show how time allocation methodology can be used to investigate certain important substantive questions. Some evidence is found to suggest that for certain applications the gain from using optimal lengths and optimal weights instead of original lengths and unit weights for subtests, as is now general practice in many testing programs, may be worthwhile. It is found that the use of the optimal time procedure with unit weights sometimes leads to results that are nearly as satisfactory as optimal times with regression weights. Finally, some evidence is found to suggest that quite different time allocations may be required for different subgroups.