Table 5
COMPLETENESS AS A FUNCTION OF ASSIGNMENT BASED ON TESTED ABILITY?.-
COMPONENT-MODIFIER SCORES
(FOR 16 EXPERIENCED PI's)
Highest Average Lowest
Target Target Score Target Score Target Score
x 61 34 0
2 74. Sl 35
3 ll 3h 15
L 88 36 0
5 62 43
6 63 28 0
7 61 Ll 19
8 61 39 L
9 TL 3o 1h
Av. 65 29 10
®The score = Stereo Zig X- 100
Expectancy Score
The group modal response for component scoring was 33% complete,
as compared with 49% for the average individual; that for the two best
was 51%, and for three random, 53%. For component-modifier scoring,
the respective completeness indexes were 72%, 39%, 43%, and 39%.
In terms of reliability in the psychometric sense, a pooling of
interpreter reports and acceptance of a consensus report as "the inter-
preter report" can be expected to improve substantially the accuracy of
the information. If the number of interpreters is sufficiently large,
the completeness of information extraction will also be improved. With
respect to completeness, however, the consensus report by two individuals
must be expected to be less complete than that by individual interpreters,
since complete overlap on all items correctly identified is not expected.
As the number of interpreters contributing to the consensus report in-
creases, the completeness score can be expected to increase.
- 36 =