Evaluating courseware development effort estimation measures and models

  • Ian Marshall

    Student thesis: Doctoral Thesis


    The aim of the work described in this thesis is to establish a theoretical and practical framework for evaluating courseware effort estimation models. Existing research into this area is thoroughly reviewed. Expert estimation, algorithmic and non-algorithmic estimation methods are evaluated in terms of well defined criteria adapted from software metrics research. Developers’ opinion on courseware effort estimation and analysis of the significant factors in improving the estimation accuracy will be discussed. A rigorous measurement paradigm to describe and assist in data collection for courseware effort estimation is defined.

    The results from twelve studies in to courseware effort estimation are used to identify seventy seven productivity adjustment factors. The productivity adjustment factors are classified into six broad groupings. The rating scale used in existing effort estimation models are presented. Nineteen critical productivity adjustment factors which affect courseware effort estimation methods are also identified.

    The implementation of the measurement paradigm and productivity adjustment factors are evaluated using courseware development data in order to prepare effort estimation models. Four courseware development case studies are used to evaluate the measurement paradigm for standardising courseware development effort measures and data collection. In addition, effort estimation models are developed using courseware development data from existing studies and a large courseware development.

    Analysis of the effort estimation models indicated that significant relationships existed between effort, learner time, key productivity adjustment factors and code measures. The effort estimation models produced disappointing results when compared against established criteria for evaluating software effort estimation models. However, despite this a number of important points were highlighted which will result in the development of better models in the future. Finally, the results and methodology used in this thesis are critically evaluated before further work and future directions for research are explored.
    Date of AwardNov 1996
    Original languageEnglish

    Cite this