Technical Report - International Military Testing Association
Technical Report - International Military Testing Association
Technical Report - International Military Testing Association
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
.\<br />
.<br />
.<br />
.<br />
”<br />
for the validutioo e8mple, item standard deviations, item varibccee,<br />
item-test correlations, item-criterion correlations, indcxea of<br />
reliability, and indexes of validity. Mr. Wry and Hr. Shirkey h&ve<br />
presented in their reports, which treat the iten relationships, 8<br />
wider di6cour6c vith an8lyeis and implications in this ares using the<br />
needed examples for illustration.<br />
The second pha6e, concerning maximization of test validity by item<br />
alalection, ir appmached by tlimiruzttng item8 from the origin;ll evaiuation<br />
te6t which contributed nothing to validity. Correlation coefficient8<br />
appear in tabular form showing the relationehips between BSMA’s and each<br />
BSHA and the criterion vith the coefficient computed to indicate the<br />
validity of the revired evaluation teet.<br />
The third phase of the sequence Fn the telt validating procedure6<br />
occurs fn a m6ximlution of test validity wfth the optiml allocation6<br />
of items acccrding to the test outline. Techniques of multiple correlation<br />
uhfch determine optinval item ellocationa have been formulated<br />
by HOr8t (1949) and Taylor (195C). Theee technique8 reflect function6<br />
In relation to the correlations between B.SM’o, their reliabflitfcs, and<br />
validities. The correla t ions, reliabilities, and validitiec are systematically<br />
ctungcd. vhen the number of item per BSX4 are altered. By<br />
increseing the less reliable Bs;yA’s. validity is further enlarged, provided<br />
the valid vrriance is not measured alao by other BWA’r. After a clueter<br />
aMly6i6 (Br*Jchter. 1954) of KiNA’s the Horst technique is applied.<br />
(This procedure in applicable currently, but oay give way in deference to<br />
the Wherry - WIncr mathod for factoring large numbcr~ of items.) In<br />
tabular form a rwmary is given of the cluster analyeie showing each<br />
ciuster, optlpal nrrmber of item6 per cluster, and the BWA’s per cluster<br />
with the valid number of items In each BRIM, the proportlon of BStIh<br />
ltem6 in each cluster, and the optiraal number of iteuis per BSHA.<br />
From a ruemary reviev of the validating procedure8 used vfth the<br />
te6t analysis, a brief description of the overall validity approach fa<br />
desirable at this point. Since urch, some six MOS terts which eVAlUStU<br />
mboot 17% of EM under the Army EES, have received validity analysis and<br />
evaluation based upon the criterion of job performsme racingr.<br />
The ssmples used rnnge from 30 to 129 and were checked to assure<br />
reprerentative groupr. The evaluation test validity coefficients<br />
ranged from .lO to .52. The CER had validity coefficients rangihg from<br />
.20 t o .51. The multiple correlation coefficient6 betveen evaluation<br />
test, CER, and the criterion ranged from .36 to .S5sdemostrating a<br />
rlightly better prediction of validity from the BT and CER combined.<br />
The corrected multiple correlation coefficients after ehrinkqe range<br />
from .34 to .51.<br />
. . . .<br />
108<br />
- .I<br />
.<br />
.