H. L. CHEN

the students with high level of test anxiety and are classified as

Group D; the subjects whose scores of test anxiety comprise the

middle third of all the scores are defined as the students with

moderate level of test anxiety and are classified as Group E; the

subjects whose scores of test anxiety comprise the bottom third

of all the scores are defined as the students with low level of

test anxiety and are classified as Group F. All the 78 subjects

are then required to take the 60-item test with moderate diffi-

culty which has been taken by Semester II subjects in investi-

gation I. The test is carried out in a fixed random item order

through conventional computerized test. Since computerized

adaptive tests which will be administered later require unidi-

mensionality, the results of 32 items which represent the mod-

erately difficult items belonging to factor 1 among the 60 items

are picked out and analyzed.

According to the test results of the 32 items, Group D is fur-

ther divided into two subgroups with no significant mean dif-

ference, they are D1 and D2 (t-test p = .649 > .05), Group E is

also further divided into two subgroups with no significant

mean difference, they are E1 and E2 (t-test p = .641 > .05), and

Group F is further divided into two subgroups with no signifi-

cant mean difference, they are F1 and F2 (t-test p = .589 > .05).

Each pair of subgroups represents subjects with similar aca-

demic ability at a certain level of test anxiety.

Next, the six subgroups are required to take tests containing

the remaining 128 items concerning factor 1 in the item bank.

The 128 items are administered to subgroups D1, E1 and F1

through a computerized adaptive test which can adjust the item

order according to individual examinee’s perceived item diffi-

culty, while the same 128 items are administered to subgroups

D2, E2 and F2 through a conventional computerized test in

which the items are arranged in a hard-to-easy order according

to item bank calibrated item difficulty. According to the results

of the tests, the moderating effects of the item order adjusted

according to individual examinee’s perceived item difficulty on

the relationship between test anxiety and test performance can

be discovered. Table 3 shows the mean differences and the

level of significance for t-test between subgroups of the same

pair.

As for Group D1 and Group D2, the results reflect that the

mean score of Group D1 which take a computerized adaptive

test and that of Group D2 which take a hard-to-easy test are

significantly different (t-test p = .009 < .05). Since Group D1

and Group D2 are the subgroups with high level of test anxiety,

it can be concluded that the item order adjusted according to

individual examinee’s perceived item difficulty may have a

significant effect on the relationship between test anxiety and

test performance as far as examinees with high test anxiety are

concerned. As for Group E1 and Group E2, the results reflect

that the mean score of Group E1 which take a computerized

adaptive test and that of Group E2 which take a hard-to-easy

test are significantly different (t-test p = .024 < .05). Since

Group E1 and Group E2 are the subgroups with moderate level

Table 3.

Mean differences and p-values observed in Investigation II .

D1-D2 E1-E2 F1-F2

Mean difference 12.1538 9.4615 9.30769

t-test p value .009 .024 .043

of test anxiety, it can be concluded that the item order adjusted

according to individual examinee’s perceived item difficulty

may have a significant effect on the relationship between test

anxiety and test performance as far as examinees with moderate

test anxiety are concerned. As for Group F1 and Group F2, the

re s ults reflect that the mean score of Group F1 whi ch take a com-

puterized adaptive test and that of Group F2 which take a hard-

to-easy test are also significantly different (t-test p = .43 < .05).

Since Group F1 and Group F2 are the subgroups with low level

of test anxiety, it can be concluded that the item order adjusted

according to individual examinee’s perceived item difficulty

may also have a significant effect on the relationship between

test anxiety and test performance as far as examinees with low

test anxiety are concerned.

Discussion

According to the results of the 60-item tests with moderate

difficulty administered to Semester II subjects and Semester III

subjects in investigation I and investigation II respectively, it

can be discovered that there is no significant difference be-

tween Semester II subjects and Semester III subjects in aca-

demic achievement (t-test p = .603 > .05). Therefore, Semester

II subjects and Semester III subjects can be regarded as two

sample populations with similar academic ability and the out-

comes from Investigation I and Investigation II can be analyzed

in a combined way. Figure 2 shows how the outcomes from the

two investigations are related with each other.

From the above diagram, some global findings about the

moderating effects of item order arranged by difficulty on the

relationship between test anxiety and test performance can be

obtained by comparing the outcomes of the two investigations.

Firstly, it can be found that both lines rise from left to right,

which demonstrates that no matter whether the item order is

arranged by item bank calibrated item difficulty or adjusted

according to individual examinee’s perceived item difficulty,

the higher test anxiety the examinee has, the more easily the

test performance of the examinee can be influenced by item

order. Secondly, according to the easy-hard: hard-easy line, it

can be found that item order has significant moderating effects

on highly-anxious and moderately-anxious subjects, but the

effect on subjects with low test anxiety is not significant; while

according to the order adjusted by perceived difficulty: hard-

easy line, it can be found that item order has significant moder-

ating effects on all subjects in three levels of test anxiety. A

vivid demonstration of the finding is that the line representing

the comparison between the easy-hard item order and the

hard-easy item order is entirely above the line representing the

comparison between the item order adjusted by perceived dif-

ficulty and the hard-easy item order, which indicates that the

item order adjusted by perceived difficulty has a greater mod-

erating effect on the relationship between test anxiety and test

performance in a whole sense.

According to the outcomes of the two investigations and the

discussion above, two conclusions can be made at least: 1) Item

order arranged by difficulty does have moderating effects on

the relationship between test anxiety and test performance. The

higher test anxiety the examinee has, the more significant the

moderating effect will be; 2) The moderating effects of the item

order adjusted according to perceived difficulty are in a whole

sense more significant than the moderating effects of the item

order arranged by item bank calibrated item difficulty.

Copyright © 2012 SciRe s . 331