09.12.2012 Views

I__. - International Military Testing Association

I__. - International Military Testing Association

I__. - 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.

intraclass correlations for many scales were zero<br />

or negative. A standardization transformation to<br />

remove scale use differences among raters was<br />

needed. For these scales, a zero rating has<br />

absolute meaning--that a task requires no skills<br />

or knowledges related to a particular scale. A<br />

standardization transformation should not change<br />

zero ratings. For this reason, the following<br />

standardization was applied to the ratings:<br />

Y ,,k = X,,, / [sum(k= 1,57) Xijk ] [Equation 41<br />

where yi,k = standardized rating for rater i on<br />

scale j and task k, and<br />

X Ilk = raw rating for rater i on scale j and<br />

task k.<br />

Figure 2 presents interrater agreement<br />

statistics for the 26 rating scales after this<br />

standardization. The ratings have acceptable<br />

interrater agreement. Three of the scales,<br />

Medical-Patient Care, Medical-Equipment<br />

Oriented, and Medical Procedures had no<br />

non-zero ratings for tasks in this occupation.<br />

Thus, meaningful intraclass correlation statistics<br />

could not be computed for these scales although<br />

all raters agreed on all ratings for these scales<br />

(zero).<br />

For modeling purposes, we augmented the<br />

training-time data file from the TDS R&D with<br />

scores on the 26 skill and knowledge scales. We<br />

used mean standardized ratings across raters for<br />

each task and scale. If the 26 scales are a useful<br />

basis for estimating training-time models, then<br />

equation 3, which uses scores on the scales along<br />

with training times to predict proficiency, should<br />

account for most of the proficiency variation<br />

accounted for by equation 2, which includes<br />

actual task identities.<br />

Results<br />

Scale ( Omega21 Rkk<br />

1. Clencal .28 .66<br />

2. Computatronal .13 .44<br />

3. Office Equipment Operatton 34 .72<br />

4. Mechanical .13 A3<br />

5. Simple Mechanrcaf .06 .23<br />

EquipmentlSystems<br />

Operatron<br />

6. Complex Mecnanicaf .Ol .06<br />

Equipment/Systems<br />

Operation<br />

7. Mecnanical-Electrical .15 .47<br />

8. Mechanfcai-Electronrc .20 .56 -.<br />

9. Elecnrcal .ll .38<br />

10. Eiewomc .20 .56<br />

11. E!ectncaf-MechanrcaJ .22 .58<br />

12. Eiectncaf-Eiectronc .13 A3<br />

13. Eiectronic-Mechantcal .19 .54<br />

14. Simple PhysrcaJ Labor .oo .oo<br />

15. Medical-Pattent Care . .<br />

16. Medical-Equipment . .<br />

Or-tented<br />

17. Medical Prccedures . .<br />

18. Simple Nontechnical .02 .I3<br />

Procedures<br />

19. Communicative-Oral .20 .55<br />

20. Communicatrve-Written .*9 .a3<br />

21. General Tasks Or .13 .e<br />

Proceaures<br />

22. Reasoning/Planning/ .05 .2?<br />

Analyzing<br />

23. Scienttfic Math Reasoning .08 .31<br />

Or Calculatrons<br />

24. Speaal Talents .05 ‘9<br />

25. Supervisory .27 24<br />

26. Training .05 .20<br />

Note:<br />

Omega Squared Is The Intrac!ass Correlation fnter-<br />

Rater Agreement: It Is Equivalent To The Fit 1. Pkk<br />

Is The Estimated Reliability For The Mean Rating From<br />

Five Raters.<br />

‘All Tasks Had Zero Ratings; Inter-Rater Agreement<br />

Statistics Are Meaningless.<br />

Figure 2. Interrater Agreement Data<br />

for the 26 Skill and Knowledge Scales.<br />

Our first modeling activity was to fit the regression model of equation 2. The R* for this<br />

model was .65, which is statistically significantly greater than zero: F(451,2255) = 9.5, p < .OOl.<br />

Next, we fit the regression model of equation 3, which replaced task identification variables<br />

with scores on the skill and knowledge scales. The R* for this model was .52, which is also<br />

statistically significant: F( l&$,2525) = 14.9, p < .OOl. If one views the skill and<br />

120

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

Saved successfully!

Ooh no, something went wrong!