Bimain 2

BIMAIN is a special form of the BILOG program designed to detect and estimate item parameter drift (DRIFT) and differential item functioning (DIF). Extremely versatile, BIMAIN is essentially a multiple-group BILOG. The groups correspond to successive cohorts of respondents when DRIFT is studied and to separate demographic groups when DIF is studied. The program employs the marginal maximum-likelihood method to estimate the effect of the groups on the location parameters of the 1-, 2- or 3-parameter logistic models under the restriction that the slopes and lower-asymptotic parameter values are the same in all groups.

In DRIFT analysis the group effects are estimated as a polynomial trend that may be set to any degree up to 5. In DIF analysis the effects are estimated as contrasts between each group and a specified reference group. BIMAIN's capability to distinguish examinee subpopulations provides correct handling of the impact of this information on item parameter estimation. BIMAIN can also be used to scale the second-stage test in a two-stage testing application. Information in the first-stage results (that are used to design respondents to the second-stage groups) is utilized by applying the posterior weights for the groups from the first-stage calibration as prior quadrature weights in the second-stage calibration.