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.

-.<br />

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

Data from this comparison sample were scored using the sanmprocedures that were used in the applicant<br />

sample. Mean item-level scores from the NROTC student sample were compared with those from<br />

the pilot-test sample in order to identify items with substantially different base rates. If an item’s mean<br />

score in the applicant group is slanted considerably more in the socially desirable direction than that of<br />

the student sample, it suggests that the item is relatively easily distorted by applicants.<br />

Refining the PEC<br />

Results from the pilot test data analyses, along with the information from the retranslation exercise,<br />

were used to revise the composition of the PEC constructs and their definitions. The inventory was then<br />

refined and shortened for future administrations. Descriptive statistics, internal consistency reliabilities,<br />

and scale score intercorrelations were computed for the final shortened scales.<br />

Evaluating the PEC<br />

Retranslation Results<br />

Results and Discussion<br />

Seventy-seven percent of the PEC items were sorted into the same predictor construct scale by five<br />

or more of the seven researchers who participated in the retranslation exercise. Seven of the remaining<br />

items were from the Unlikely Virtues (response validity) scale. It is not particularly surprising that some<br />

researchers sorted the Unlikely Virtues items into the construct categories. The Unlikely Virtues items<br />

were specifically written to resemble the eight original construct categories (so they would be subtle).<br />

The fact that some of the judges mistakenly sorted the Unlikely Virtues items into the construct categories<br />

suggests that the items are indeed subtle. In general, however, there was good agreement among the<br />

researchers concerning the placement of PEC items on constructs.<br />

Pilot-Test Results<br />

Frequency counts revealed that the vast majority of the PEC items had an adequate spread of responses<br />

across the response alternatives. Only a few items had response distributions that were considered<br />

unacceptable (e.g., over 90 percent of the respondents chose the most desirable response alternative).<br />

However, for some items the response distributions were much better than for others. This information<br />

was taken into account in refining the PEC, particularly in making decisions concerning which items to<br />

drop.<br />

The item-level intercorrelations were factor analyzed, and rotated principal factor solutions containing<br />

from 2 to 12 factors were examined. Based on a parallel analysis (Montanelli & Humphreys, 1976),<br />

the amount of variance accounted for by each factor, and the interpretability of the solutions, the eight<br />

factor solution was selected for further consideration,<br />

The amount of overlap between the results of the retranslation exercise and the factor analysis was<br />

encouraging. Items from the Leadership Orientation scale defined a factor, and nearly all of the items<br />

that were retranslated into this scale had their highest loading on that factor (8 of 11). Similarly, Organization<br />

and Planning and NROTC/<strong>Military</strong> Interest and Motivation also defined their own factors.<br />

Achievement Motivation defined a factor, but most of the Responsibility (7 or 12) items also loaded on<br />

this factor. Dominance and Team Orientation each defined a factor, and the Sociability items were split<br />

between these two factors, with the Sociability items involving friendliness loading on the Team Orientation<br />

factor and those involving talkativeness and assertiveness loading on the Dominance factor. Clearly the<br />

retranslation and the factor analysis results converged on very similar sets of constructs.<br />

500

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

Saved successfully!

Ooh no, something went wrong!