27.03.2014 Views

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SEKE 2012 Proceedings - Knowledge Systems Institute

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

B. Instrumentation<br />

The experiment was supported by a set of artifacts: consent<br />

forms, software product lin e specification, the SPLIT<br />

documentation for the group that would execute them ,<br />

worksheet to support defects and follow up questionnaire. The<br />

software product line specification is composed by the<br />

requirement specification, product map and feature model.<br />

The software requirement artifact is a software product line<br />

specification for a set of twitter client products with 13 features<br />

from popular tw itter clients having 6 m andatory features, 3<br />

optional features and 4 alternative features. It generates six<br />

different products sharing the common features and each of<br />

them having distinguished featur es. It describes each feature<br />

for the software product line detailing its relationship to others<br />

features, classifying it as mandatory, optional or alternative and<br />

enumerating the products in which this feature would be<br />

available.<br />

The product map for the twitter client software product line<br />

listed all the features available in softw are requirement<br />

document and associates it to the product in which they are<br />

available. It enumerates the six products from the twitter client<br />

software product line and maps the features for each product.<br />

The feature m odel for the software product line has been<br />

created according to the softw are requirements for the tw itter<br />

client and it would generate all the products, which are<br />

specified in the product map. The feature model used in this<br />

experiment has been seeded with defects by the researchers<br />

based on deficiencies from feat ure models listed in [7]. It<br />

contains defects from mode ling the feature model and<br />

information inconsistencies from the softw are requirements<br />

and the product map.<br />

C. The Experiment Design<br />

The subjects were divided in two groups, which would<br />

inspect the same software pr oduct line model example: the<br />

group A would evaluate using a defect type based inspection<br />

approach and the group B would use the set of techniques<br />

proposed in this paper. Sin ce no student had a previous<br />

experience with SPL models, the subjects were assigned to<br />

each technique using completely randomized design. Each<br />

group was composed by 10 senior-level undergraduate students<br />

chosen by convenience from Analysis and Design class in<br />

Information System and Computer Science courses at Federal<br />

University of Amazonas, Brazil.<br />

D. Preparation<br />

The subjects signed a consent form and they had a tutorial<br />

about Software Product Line. It addressed an overview on<br />

Software Product Line concepts and specifications including<br />

feature models and product map. After the tutorial, the subjects<br />

have been trained for the experiment execution. As the<br />

classroom was divided in two groups, two different trainings<br />

were prepared. The Group A has been trained in software<br />

model inspection techniques. This training addressed the<br />

objectives of model inspection for software product lines. It<br />

described the defect types, which could be found by the<br />

inspectors when conducting the inspection in this study. The<br />

group B has been trained in the set of techniques proposed in<br />

the SPLIT. The two trainings ha d different instructors due to<br />

the fact that they occurred at the same time for avoiding<br />

communication between the two groups. However both<br />

materials have been prepared by the two instructors for having<br />

balanced knowledge about softw are inspection for m itigating<br />

the bias of prior knowledge of the subjects.<br />

E. Execution<br />

The experiment have been executed in a lim ited two hours’<br />

time box and occurred at the same time for both groups. The<br />

subjects were gathered in same room and then divided<br />

randomly in groups. Group A and B have executed the<br />

experiment in different rooms for avoiding any communication<br />

between subjects from different groups.<br />

V. RESULTS<br />

At the end of the experim ent, the defect list form and the<br />

evaluation questionnaire about the proposed set of techniques<br />

were retrieved for the experim ent analysis. The num ber of<br />

defects found per subject for each inspection is presen ted in<br />

Table IV and Table V.<br />

TABLE IV.<br />

NUMBER OF DEFECTS FOUND IN DEFECT TYPE BASED<br />

INSPECTION APPROACH<br />

Subject<br />

Number of Defects Found<br />

Subject 01 4<br />

Subject 02 9<br />

Subject 03 7<br />

Subject 04 8<br />

Subject 05 4<br />

Subject 06 9<br />

Subject 07 1<br />

Subject 08 3<br />

Subject 09 7<br />

Subject 10 7<br />

TABLE V.<br />

NUMBER OF DEFECTS FOUND USING SPLIT<br />

Subject<br />

Number of Defects Found<br />

Subject 11 10<br />

Subject 12 11<br />

Subject 13 12<br />

Subject 14 8<br />

Subject 15 10<br />

Subject 16 13<br />

Subject 17 8<br />

Subject 18 10<br />

Subject 19 8<br />

Subject 20 10<br />

660

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!