An Empirical Study On Integrating Analytical Quality Assurance Into Pair Programming

Posted on September 14th, 2006 by gernot.
Categories: Publications, Studies.
Tags: , ,

Authors:

  • Dietmar Winkler
  • Ramona Varvaroi
  • Gernot Goluch
  • Stefan Biffl

Conference:

ISESE2006

Abstract:

Defects in software design can have major impacts on product quality, project duration and budget. Analytical quality assurance tasks, like software inspection and testing, help detect deviations.
Recently, Pair Programming has been introduced as constructive approach for agile code construction that includes implicit quality assurance approaches, e.g., continuous defect detection, but without active guidance. Several empirical studies recommend active guidance for more efficient and effective defect detection.
In this paper we propose an extension of pair programming that integrates best-practice inspection and test case generation
approaches in order to improve defect detection in software products. We discuss the concept of a controlled experiment to empirically evaluate the proposed pair programming approach.

0 comments.

An Empirical investigation on the Visualization of Temporal Uncertainties in Software Engineering Project Planning

Posted on December 21st, 2005 by gernot.
Categories: Publications, Studies.
Tags: ,

Institute of Software Technology and Interactive Systems, Vienna University of Technology, A-1040 Vienna, Austria,

Authors:

  • Stefan Biffl
  • Bettina Thurnher
  • Gernot Goluch
  • Dietmar Winkler
  • Wolfgang Aigner
  • Silvia Miksch

Conference:

ISESE 2005

Abstract:

The success of software projects depends on the ability of a human planner to understand the relationships of tasks and their temporal uncertainty and hence the visualization thereof. In this paper we report on an empirical study that compares the performance of two techniques to visualize task relationships and temporal uncertainties: traditional “best-practice” PERT charts and recently introduced PlanningLines. Main results of the study are: (a) while PERT charts are well suited for reading single attributes, PlanningLines better support users in judging temporal task uncertainty; (b) both experiment rounds shows consistent results regarding the strengths and limitations of the techniques. Overall, these results suggest that a combination of PERT charts and PlanningLines has the potential to significantly improve the planning support of project managers and software engineers.

0 comments.

Usability Test Planning Lines

Posted on February 16th, 2005 by gernot.
Categories: Studies.
Tags: ,

back to Work >> back to Other

Goal of the study was to compare the performance of ‘Planning Line’ glyphs (PL) with PERT charts.
The reason for the decision to compare PL with PERT lies in the ability of PERT to display uncertainties, related to the ending, beginning and duration of a certain project task. Exactly this ability is one of the main advantages, the ‘Planning Lines’ graph provides. The reason for not using similar graphs, like the GANT graph, is that they are not able to display uncertainties in time. Following this the comparison with PL would not be very expressive.

1. Subjects

The subjects are graduate students of informatics and business informatics in a usability engineering workshop.
The Austrian university system allows students to organize their studies in a very individual way. This enables them to determine their own schedule and to work part-time during their studies. Therefore the subjects participating in this workshop exhibit rather heterogeneous knowledge and experience levels.
We provided a tutorial, that briefly repeated how to use PERT, a method known by most participants, and introduced the new method ‘Planning Lines’, to guarantee the minimal common level of knowledge for the experiment.

2. Number of Subjects

In December 2004, 48 students of the workshop ‘Usability Engineering’ participated in the experiment.
From the experiment preparation team four members supervised the experiment and answered questions concerning the experiment or questionnaire.

3. Experiment Objects and Procedures

This section provides a short overview of the experiment objects, which were used in our study.
The experiment participants received the following experiment materials

  • 1. Background Questionnaires: A one-sided questionnaire was provided at the beginning of the experiment. Participants were asked to give general information (Name, Age…) and specific information about their experience with PERT and similar graphs.
  • 2. Answering sheets for task solutions
    • a. Part A: This part contained a three sided answering sheet for questions and tasks, concerning the usage of PL or PERT. Four different versions of this part were available, differing in treatment (PL, PERT) and data set (1, 2), and were randomly handed out to the subjects.
    • b. Part B: This part contained a project plan and a five sided answering sheet for questions, which were based specific tasks in the project plan. Four different versions of this part were available, differing in treatment (PL, PERT) and data set (1, 2), and were randomly handed out to the subjects.
    • c. Part C: On an on-sided answering sheet participants had to draw a graph, based on textual task information. Once again four versions were available, differing in treatment (PL, PERT) and data set (1, 2), and were randomly handed out to the subjects.
  • 3. Feedback Questionnaire: At the end of the experiment every subject had the possibility to give his or her feedback to the PL graph.
    Before handing out the experiment material, in the order listed above, a short tutorial, that briefly repeated how to use PERT, a method known by most participants, and introduced the new method ‘Planning Lines’, to guarantee the minimal common level of knowledge for the experiment, was held by one of the experiment design members.
    Afterwards the experiment material was handed out. There was a time limit of 45 minutes to finish the given questionnaires and answering sheets. Additional to this subjects were asked to note down the current time before start and after finishing a part of the answering sheet. This was done to measure the time every participant needed to solve the given questions and tasks (in some hypotheses we refer to the time a subject needs to finish a task).

Finally subjects had the possibility to give feedback to the PL graph in a one-sided questionnaire.

Comments Off