Full-Information Item Factor Analysis of Tests Forms From the ASVAB (Armed Services Vocational Aptitude Battery) CAT (Computerized Adaptive Test) Pool

摘要

Data for nine tests in the computerized adaptive test pool for the Armed Services Vocational Aptitude Battery were subjected to item factor analysis by the marginal maximum likelihood method. Because each of the tests was represented by three distinct forms to which items had been randomly assigned, the stability of the results could be examined by comparing independent analyses. Certain items gave clear and generally interpertable factor solutions on all tests. The statistical significance of each principal factor after the first was tested by means of the large sample chi-square statistic corresponding to the increase in marginal maximum likelihood as successive factors were added. To allow for possible cluster effects in the sampling of respondents, a design effect of two was assumed when computing these test statistics. Apart from minor factors associated with doublets and occasional item format effects, the main factors were all interpertable. Most notably, General Science showed a distinction between business arithmetic and other calculation problems. Shop Information showed a clear distinction between Metal Shop vs. Carpentry and General Contracting. The most interesting finding was a clearly significant two factor solution for Word Knowledge. One of these factors appears to represent vocabulary drawn from literary sources and recreational reading, and other from television and the news media. Implications of these findings for vocational testing are discussed.