Hardware Reference
In-Depth Information
Table 1.
Results from annotations experiment
Row Q
Group Type
Errors
Correct
Missing
1
Q1 B
Annotated
7.33
12
14
2
Q1 A
Plain
13
14
12
3
Q1 -
Difference
55.74% (15.38%) (15.38%)
4
Q2 A
Annotated
4.67
14
7
5
Q2 B
Plain
9.67
10.67
10.33
6
Q2 -
Difference
69.77% 27.03%
38.46%
7
Q3 B
Annotated
9.67
9.67
3.33
8
Q3 A
Plain
11.67
9
4
9
Q3 -
Difference
18.75% 7.14%
18.18%
10
Q4 A
Annotated
11.5
14
6
11
Q4 B
Plain
14
8.67
11.33
12
Q4 -
Difference
19.61% 47.06%
61.54%
13
-
-
Average Difference 40.97% 16.46%
25.70%
Further, the percentage difference (55.74%) between Group A and Group B
is represented in row 3. This is obtained from as follows:
Row
|Row
− Row
|
/
Row
Row
/
.
3=((
1
2
)
(
1+
2)
2)
100
(1)
For example, in the case of the obtaining the percentage difference of Errors:
55
.
74% = (
|
7
.
33
13
|
)
/
((7
.
33 + 13)
/
2)
100
(2)
In the case of a worse performance when given annotations, such a result has
been enclosed with parenthesis in Table 1. This pattern continues for each ques-
tion given to respondents. The final row, Row 13, contains the overall percent-
age difference; these results included the cases of decreased performance in Row
3 as negative numbers. In each case, the occurrence of Errors is reduced for
the annotated versions. This holds across both groups even with a pattern of
Group A taking less time on average compared to Group B. For example, the
figure of 55.74% in row 3 indicates that there were 55.74% less errors identi-
fied in the annotated version. This means responses with a lower incidence of
Over-Specification occurred when respondents were provided with annotations.
In Question 2 to Question 4, the average number of Correct Elements for the
annotated version is greater than that for the non-annotated version. A similar
reduction in the number of Missing Elements occurred. For example, 27.03%,
Correct in Row 6 means that there were 27.03% more elements identified by the
group given annotations. Similarly, 38.46%, Missing in Row 7 means that there
were 38.46% less missing elements identified by the group given annotations.
As with Error Rates ,thenumberof Correct Elements achieved by respondents
appears unrelated to the amount of time spent. However, the effect of annota-
tions on Correct Elements and Missing Elements wassmallerthanonthe Error
Rates , therefore annotations had less of an impact on Under-Specification .
Search WWH ::




Custom Search