| Equating is to establish an effective equivalence between scores on two testforms which measure the same construct or the same psychological trait. The goal ofequating is to compare the scores of different tests. The non-equivalent groups withanchor test (NEAT) design is one of the most flexible tools available for equating tests.It is a widely belief that an anchor test used in equating should be a miniature version(or minitest) of the tests to be equated; More specifically, the anchor test should beproportionally representative of the two tests in content and statistical characteristics.Currently, most operational testing programs that use the NEAT design employ aminitest as the anchor; to ensure statistical representativeness, the usual practice is tomake sure that the mean and spread of the item difficulties of the anchor test areroughly equal to those of the total tests. Sinharay and Holland(2006) suggest that wecan relax the spread of the item difficulties. This paper examines this belief, especiallyregarding statistical characteristics.This paper using R-2.15langurage Software to achieve simulation study theeffect of Anchor tests with a spread of item difficulties on test equating. Using threeequating method (chain equipercentile equating, Levine method and Tucker method)and non-equivalent groups with anchor test to compare Several types of anchor testshow to effect test equating。Anchor tests with a spread of item difficulties less thanthat of a total test seem to perform as well as a minitest with respect to equating biasand equating standard error. Hence, the results demonstrate that requiring an anchortest to the same statistical characteristics of the total test may be toorestrictive.Furthermore, a minitest has to include very difficult or very easy items toensure adequate spread of item difficulties, which can be problematic as such itemsare usually scarce. If relaxes the requirement on the spread of the item difficultiescould be more operationally convenient. |
PDF Full Text Request