Constructing A Multidimensional Aptitude Test (MAT)

Objective: The Multidimensional Aptitude Test (MAT) was developed to provide an objective measure of general aptitude in order to administer personnel-selection and placement for certain occupational groups such as Armed Police Force. The present study was conducted to construct the scale, analyze the items and examine its reliability and validity.Method: We formed the structure of MAT through job analysis including expert consultation and opened questionnaire et al. The MAT consisted of four sub-scales, namely, abstract thinking, spatial ability, attention & speed, motor stability, There were totally fifteen sub-tests. 8 sub-tests came form Changsha-Anshan Intelligence Test in Group (CAITG), and the rest were developed creatively. Items were screened through testing in advance. A total of 991 Armed Police Force men were tested in the formal study. Otherwise, 45 subjects take the MAT re-testing. 33 subjects were tested by WAIS, and 303 subjects' Abilities were rated by their leaders. The military training scores of 136 subjects were collected. Reliability testing included test-retest and Cronbach's a coefficients. Validity testing included construct-related validity and criterion-related validity. The construct-related validity was examined by exploratory factor analysis (EFA)and confirmatory factor analysis (CFA). The criterion-related validity was examined by group differentiation, training scores, leader's rating, and WA1S respectively.Results: 1. Item analysis: The difficulty of sub-tests ranged from 0.38 to 0.60, while the average difficulty of the whole scale was 0.50. The discrimination of 85% items was over 0.2. 2. Reliability: The Spearman test-retest correlations for each sub-scales ranged from 0.73 to 0.88, with a total score test-retest reliability of 0.79. Cronbach's a coefficients for each sub-scales ranged from 0.53 to 0.75, with a whole homogeneity reliability of 0.80. 3. Validity: There were higher correlations between the total score and three sub-scales: abstract thinking, spatial ability, attention & speed (0.68-0.80), while lower correlations between the total score and motor stability (0.29). There were four main factors by EFA, accounting for54.28% of whole variance. Fit indexes by CFA, CFI and NNFI were over 0.95, RMSEA was below 0.05, CMIN / DF was below 2. There were significant differences between jobs and educational levels (p<0.01 or P<0.05) . Significant correlations were obtained betweenMAT and exam scores, leader's rating, and WAIS respectively.Conclusion: Although further examination of the .psychometric properties of MAT is warranted, the present study has proved that 1. The structure ofMAT is reasonable and innovative. 2. The quality of the items meets the need of psychometrics. 3. The reliability is acceptable. 4. Theconstruct-related validity is reasonable, and theoretic model provides a good fit to the data. 5. Criterion-related validity is supported by some examinations.