Research for Statistics

Four logit specification tests for the case of grouped data have been developed. The first test compares the covariance matrix of the regression residuals to the matrix expected under the logit model. The second and third tests directly address the zero restrictions and cross-equation equality restrictions associated with the Independence of Irrelevant Alternatives (IIA) property. The fourth test embeds the logit model within a grouped data version of the nested logit model, which does not possess the IIA property. The logit model was marginally rejected in one data set, and strongly rejected in the other two data sets.

Differences in earnings of males and females have been attributed to differences in personal characteristics, fields of specialization, and years of professional experience. This study focuses on scientists and engineers, seeking to determine the extent to which differences in measured characteristics explain differences in earnings. In addition to documenting gender differences in the earnings of new entrants and experienced personnel, the study investigates gender differences in the choice of college major and in the transition from school to work.

The the feasibility of validating ASVAB enlistment standards against job performance is evaluated. Hands-on and written proficiency tests were developed for three Marine Corps skills--Ground Radio Repair, Automotive Mechanic, and Infantry Rifleman--for use as measures of job performance. In addition, grades in skill training courses were also evaluated as possible measures of job performance. The ASVAB was shown to be a valid predictor of job performance. All three measures--hands-on tests, written tests, and training grades--were generally consistent measures of performance. A preliminary set of ASVAB qualification standards for assigning recruits to these three skills was computed using the hands-on and written tests as the criterion measure. The ASVAB standards derived from this analysis are similiar to the standards based on the traditional criterion measure of training-course grades. We conclude that validating ASVAB enlistment standards against job performance appears to be feasible. Although job performance tests can be used for this purpose, they are costly to develop and administer. Training grades, which are routinely available, may serve as a satisfactory and economical proxy for them in many skills.

This report examines the validity of the Armed Services Vocational Aptitude Battery (ASVAB) forms 6 and 7. Validity in this analysis is the correlation between ASVAB test scores and subsequent performance in military training courses. Recruits are assigned to specific military training based, in part, on their scores on subgroups of tests (aptitude composites) contained in the ASVAB. We determined the most appropriate aptitude composite, and minimum acceptable score on that composite, for assignment to each training course. ASVAB forms 6 and 7 (the source of test score data for this analyses) are compared with the recently introduced ASVAB forms 8, 9, and 10. Based on this comparison we consider the results of our validity analyses of ASVAB forms 6 and 7 to be applicable for recruit assignment using ASVAB forms 8, 9, and 10.