The Sheffield Software Engineering Observatory Archive: Six Years of Empirical Data Collected from 73 Complete Projects Export

@techreport{citeulike:4039255, abstract = {Background: Software development processes and techniques are used in a development context; controlled experiments typically trade less realistic contexts for repeatability. This paper describes an alterative method which includes a realistic context in a quasi-experiment.Objective: To understand how software development processes behave in context we have conducted a series of experiments primarily comparing extreme programming and plan driven approaches. Method: A quasi experiment was used that utilised a randomised complete block design [1] with a pre-test. Over the period of six years there were 20 blocks (projects) and 74 subjects (teams) assigned to the blocks. In total there were 322 students studied.Results: The experimental approach only uncovers small differences between the teams using different treatments. This is related to the variance in tasks undertaken and the fidelity to which the treatment is applied. Conclusion: This type of experiment can help to assess the generalizability of a technique by observing how it is used in the field. Due to the variations fidelity it is not an appropriate method to check the definitive benefits of a technique when there are no other pressures. Example: A team of 6 six students use Extreme Programming techniques to develop a software product in PHP working with a client. The client knows that he wants a document management system, with several classes of users but the team must help him identify his remaining needs.}, address = {Sheffield, UK}, author = {Thomson, Chris and Holcombe, Mike}, citeulike-article-id = {4039255}, institution = {Department of Computer Science, University of Sheffield}, number = {CS-09-01}, posted-at = {2009-02-12 15:16:18}, priority = {2}, title = {The Sheffield Software Engineering Observatory Archive: Six Years of Empirical Data Collected from 73 Complete Projects}, year = {2009} }

TY - RPRT ID - citeulike:4039255 L3 - citeulike-article-id:4039255 N2 - Background: Software development processes and techniques are used in a development context; controlled experiments typically trade less realistic contexts for repeatability. This paper describes an alterative method which includes a realistic context in a quasi-experiment.Objective: To understand how software development processes behave in context we have conducted a series of experiments primarily comparing extreme programming and plan driven approaches. Method: A quasi experiment was used that utilised a randomised complete block design [1] with a pre-test. Over the period of six years there were 20 blocks (projects) and 74 subjects (teams) assigned to the blocks. In total there were 322 students studied.Results: The experimental approach only uncovers small differences between the teams using different treatments. This is related to the variance in tasks undertaken and the fidelity to which the treatment is applied. Conclusion: This type of experiment can help to assess the generalizability of a technique by observing how it is used in the field. Due to the variations fidelity it is not an appropriate method to check the definitive benefits of a technique when there are no other pressures. Example: A team of 6 six students use Extreme Programming techniques to develop a software product in PHP working with a client. The client knows that he wants a document management system, with several classes of users but the team must help him identify his remaining needs. IS - CS-09-01 AD - Sheffield, UK TI - The Sheffield Software Engineering Observatory Archive: Six Years of Empirical Data Collected from 73 Complete Projects AU - Thomson, Chris AU - Holcombe, Mike PY - 2009/// ER -