Software Reading Techniques by Unknown

Author:Unknown
Language: eng
Format: epub

Figure 4-5. The p-q phase diagram

Chapter 4 ■ SCenario-BaSed reading teChniqueS

Contradicting experimental results can be understood within this theoretical framework.

1. We first point out that this simple relationship is obtained under the assumption that reading techniques, perspective-based reading, and any other techniques that perspective-based reading is compared to can uncover all defects.

In reality, experimental conditions may not satisfy this

assumption. We didn’t see any analysis to confirm that all defects could be uncovered, in theory, using the investigated reading techniques.

2. In practice, we also find that some defects are easy to uncover, while others are not so. That is, each defect has a different probability to be detected. The analysis assumes a uniform detection probability in the defect groups (the perspective targeted or not targeted, or the entire defects).

3. When designing reading perspectives for their targeted defects, we purposely minimize the defects overlap, i.e., C region in the diagram. This turns out to be a very difficult

objective to achieve. There is no discussion from experimental designers regarding how the defect overlapping region looked like in their experiment. The existence of a common region favors perspective-based reading, however (see the second term in D pbr -D cbr ).

4. The defect detection probability, p, q, and s, strongly depends on an individual’s experience and skill. In all

published experiments, there was no explicit estimate of

those probabilities. Hence, when perspective-based reading outperforms other reading techniques, we are not clear

why and how we attribute its success; when it does not

outperform, we don’t know why it is so.

5. For future experiments, we suggest researchers collect or estimate those data so that the results can be analyzed

quantitatively and the success can be properly attributed to. Let’s examine the common defects detected by two reading perspectives. This is not

the common region C in the diagram, as the defects not targeted by a reading scenario are detected at a probability q. In the two-person team, on average, the common defects reported by both team members are, respectively, for perspective-based reading and checklist-based reading:

DA pbrc =+ () Bp ⋅+⋅ qCp 2 DA cbrc =++ () BC ⋅ s 2

Chapter 4 ■ SCenario-BaSed reading teChniqueS

With those two quantities, we can rewrite the difference as ∆= () AB + ⋅+−() pq 22 sC +⋅− () ps +− () DD cbrc pbrc

In literature it was believed that the effectiveness of perspective-based reading was due to the reduction of overlapping defect detection (i.e. reduced value of D cpbr ). We argue this is not necessarily the case. Suppose we have perfectly focused reading scenarios that do not detect any defects the scenarios are not targeting, i.e., q=0. In that case the overlapping defects detected by perspective-based reading is D cpbr =0 when C=0 or A+B>>C, and of course, to ensure ∆>0, p has to satisfy some condition, which would be p>2s-s 2. Now suppose for a given p and q, two reading techniques give the same performance, i.e., ∆=0. Now holding q constant, we improve the reading scenarios such that all targeted defects are detected with probability p=1.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.