15.11.2012 Views

Technical Report - International Military Testing Association

Technical Report - International Military Testing Association

Technical Report - International Military Testing Association

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Approaches to Improved Measurement:<br />

Research in Progress<br />

RCMLD K, GOODNIGHT (<br />

US Army Enlfated Evaluation Center<br />

Throughout the confer&e we have heard and discussed numerous topic6<br />

epanning the realm of test immrovement and development. In every cane<br />

each method used, the reaulto obtained, and the utilization of these<br />

re6ults are all bared on reresrch. Reeearch is the guideline to BUECC~U<br />

in almoet any endeavor. Tnncrefore, I am plcaeed to present to you several<br />

of the more important reeearch projects which have been completed, as wel.1<br />

68 nome currently in progrerit, at the US Army Enlirted Evaluation Center.<br />

The first project, completed oeveral years ago, ir A Comparison of<br />

Seven Methods of Computing Total Test Reliability fromma Single Terr<br />

Adminietration. Ttdt reliability (that is--a determination of teat conriatency)<br />

is evaluated by Lnvtstigating tht relationship of the praiictor<br />

to itrtlf, or in other words, it is-the relationship of the ranking of<br />

scores on one administration with the ranking on a aubsequant adminlstratfon.<br />

The rsliabilfty of a test can generally be estimated adequately<br />

fran only one administration of a test.<br />

For tach MOS Evaluation Teat a reliability coefficient is computed<br />

via the Kuder-ttfchardoori Formula 20. However, it ~68 imperative tc know<br />

if thie wa6 the most accuratt indication of the consirtancy with which<br />

the tert was measuring job proficiency. Thtrefore, this project was<br />

designed to determine which of the odd-even, K-R 20 on the total group,<br />

K-R 20 on 54% of the group, Hoyt’s Ana1ysi.e of VerCanct. Horat Maximum.<br />

Horet Corrected, and Cleman’r Maximum reliabili*.v mearurement methods wan<br />

super for. The result8 showed the K-R 20 relfability sethod on the total<br />

group io the meet appropriate for the MOS Evaluation Te6ta, thus supporting<br />

It8 uragt by thL Enlisted Evaluation Center (EEC).<br />

Thi8 study was completed in February 1963, and prcaently another project,<br />

A Canpariron of Fivt Hethods of Computing Total Ttrt Reliability frus<br />

a Single Test Administration, ir under way. This 18 a replication of the<br />

earlier study with some minor variation6 tc further verify EECI currant<br />

procedure8 in tt6t reliability estimation. Also, another reason for con=<br />

ducting this study lies in the proposed technical rtcommendationr which weft<br />

presented at the 1964 American Psychological <strong>Association</strong> meeting&; their<br />

rscamntndationr indicate that the K-R 21 ia more appropriate to use then the<br />

K-R 20 under situations such as those at EEC. Therefore, in this rtudy five<br />

methods of reliability eatimation-- K-R 20, K-R 21, odd-even, Hoyt’8 Analysis<br />

of Variance, and Uorst’o Maximum--are beir.g compared to determine which is<br />

the better method of mearuring te6t cons,etency in view of APA’s rtccmmendations<br />

and to rtudy the effect8 of aample sire and MOS rkill level on the<br />

varioue reliability coefficlentr. Results are not available thus far.<br />

116<br />

-_ ._.....- .--- .---. --.<br />

c<br />

_I__...

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!