03072cam a22004692u 4500001000800000005001700008007001400025008004100039035002000080037001900100040002700119088001600146091000800162100002100170245013500191260006800326300001000394500010400404500005100508520130100559521002401860530007901884533008101963650002702044650003302071650002702104650002902131650002802160650003502188650003102223650003302254650002802287650002202315650002302337653002702360653002502387655003002412655003302442710004802475856003702523984004202560542979220181019152035.0he u||024||||080220s1983 xxu ||| bt ||| | eng d 9(ericd)ED234056 aED234056bERIC aericdbengcericddMvI aETS-RR-83-3 amfm1 aBoldt, Robert F.10aStatus of Research on Item Content and Differential Performance on Tests Used in Higher Educationh[microform] /cRobert F. Boldt. a[Washington, D.C.] :bDistributed by ERIC Clearinghouse,c1983. a42 p. aAvailability: Educational Testing Service, Research Publications, R116, Princeton, NJ 08541.5ericd aEducational level discussed: Higher Education. a"Outlier studies" identify items for which extreme differences in performance by contrasting groups occur; these extreme items are the "outliers" referred to. Review of the studies conducted on tests receiving major use in higher education reveals that though one cannot make a priori classifications of outliers with confidence, one can with reasonable confidence predict the relatively advantaged group for many verbal items if they subsequently prove to be outliers as follows: aesthetic-philosophical, human relations or female oriented content relatively favors females as opposed to males; practical affairs, science or male oriented content relatively favors males as opposed to females; science content relatively favors whites as opposed to blacks. For test content that varies in the degree of relatedness to minorities, one would predict a relative advantage for those outliers that are most related to minorities. The magnitude of the differences found is not large; perhaps larger differences would be found if classifications other than race and sex, which are the most common, were used. It has been found that differences in cultural or national origin produce larger discrepancies in item difficulty than differences in race or sex of essentially native American groups. (Author)8 aResearchers.bericd aMay also be available online. Address as at 14/8/18:uhttps://eric.ed.gov/ aMicrofiche.b[Washington D.C.]:cERIC Clearinghouseemicrofiches : positive.07aClassification.2ericd07aCultural Differences.2ericd07aGroup Behavior.2ericd07aHigher Education.2ericd17aMinority Groups.2ericd17aPredictive Measurement.2ericd07aRacial Differences.2ericd17aResearch Methodology.2ericd07aSex Differences.2ericd17aTest Bias.2ericd17aTest Items.2ericd1 aOutliersaTest Content0 aDiscrepancy Analysis 7aReports, Research.2ericd 7aInformation Analyses.2ericd2 aEducational Testing Service, Princeton, NJ.41uhttps://eric.ed.gov/?id=ED234056 aANLcmc 2253 ED234056d77000000240075