This research is to find the right number of questions per category and an ideal set of conditions for
computing relative significance value, thus providing an alternative as a better originality assessment tool.
61 elementary school children took the TTCT figural A and were each reviewed by both regular scoring
version and steamlined version. Here is the summary of meaningful outcome. First, in each scoring
method, relative influences of factors affecting the subjects’ score were analyzed and it was shown that
each testing method had in common a high error variance from interaction between the subjects and
items. Also, subject error variance was shown relatively higher than items error variance. Second, when
computing the total score, steamed version was short of reliability on generalizability coefficient whereas
regular version showed a reliable result above .8. Steamlined version was inadequate to produce ideal
condition for relative significance and regular version has shown adequate reliability under current relative
significance value, .18 : .82, as well as under .2 : .8, a weighted value on activity 2, which showed high
error variance. Lastly, each scoring method was reanalyzed under different combinations of question per
category and relative significance. Steamlined version didn’t satisfy the reliability coefficient value under
any combinations and regular version showed acceptable reliability value when number of questions in
activity 3 was reduced by up to 10. At that point, optimal relative significance on activity 2 and 3 were
.2 : .8. In summary, optimal test condition that satisfies both reliability and efficiency was produced only
in regular scoring version. Also, when performing the test with only activity 2, perhaps due to lack of
time, results of the research underscored the need to find solution to secure reliability.