Statistical Project Control

Statistical Project Control

Statistical Project Control consists of a set of techniques to improve the accuracy and reliability of project control methods using Earned Value Management metrics and simple or advanced statistical methods. In the chapters below, you find information on our papers and additional material to download to replicate the studies we have published in the literature.

Paper 1. Using Tolerance limits as triggers for actions

This paper entitled "Setting tolerance limits for statistical project control using earned value management" has been published in Omega and more information can be found here. 

Paper 2. A multivariate approach for top-down project control

This paper "A multivariate approach to top down project control using earned value management" is published in Decision Support Systems. This paper contains 4 appendices, which can be downloaded here. The four appendices can be read as supportive material to the paper, and consist of the four following sections:

  • Appendix A. Calculating principal components: A short tutorial
  • Appendix B. Geometrical interpretation of principal component analysis
  • Appendix C. Tolerance limits for multivariate schedule control metrics
  • Appendix D. Illustrative example

Paper 3. Classifying empirical project data

In the paper "Empirical perspective on activity durations for project management simulation studies" written by Colin and Vanhoucke (2015), a classification methodology has been presented to classify empirical data in different classes such that they can be used in simulation studies. The classification method makes use of proven methods from literature and statistical analyses and is implemented in R. In the appendix of this paper, a short summary is given how these R files should be used. The files that must be downloaded are:

  • Colin_and_Vanhoucke_2015_template: The R template to run the method proposed in the paper
  • Batselier_Vanhoucke_2015.xlsx: The data for the 24 projects, including baseline and real durations
  • Trietsch_etal_2012.R: The R syntax for the four-step procedure of Trietsch et al. (2012)
  • Grouping_tests.R: The R syntax for the classification as presented in the paper
  • Looney_and_Gulledge_1985.txt: The table that is input for the R files
  • Casestudy.Rdata: Additional data for the case study extension in the paper
These files can be downloaded here and must used as follows:
 
Preamble: The preamble of the R template includes statements to clear the workspace, to load a package to read excel files and to define some subsidiary functions. In addition, the code is sourced to load their functions into the workspace.
 
Data: The activity-level data from the projects in the dataset of Batselier and Vanhoucke (2015) are read from the excel file (Batselier_Vanhoucke_2015.xlsx). This excel file contains the planned and actual durations for all activities in the 24 projects that are suited for the analysis.
 
Four-step procedure of Trietsch et al. (2012): The four-step procedure of Trietsch et al. (2012) is applied to these 24 projects. If the Parkinson distribution with a lognormal core is not confirmed for a certain project, the four-step procedure returns NA as output. If however, the distributional model for a project is confirmed, the function returns a list for that project which includes:
  • The step (P0,P1,P1.1,P1.2) in which the hypothesis was confirmed
  • The estimates for the mean and the standard deviation
  • The p value
  • A list of the values in the empirical relative distribution
 
Classification: The classification of table 3 was produced from the grouping of the projects with respect to their sˆ and mˆ values. This grouping can be reproduced through the application of Levene’s and Welsch’s tests to the output of the four-step procedure. First the statistical interpretation of sˆ can be reproduced, and next, the statistical interpretation of mˆ can be repeated. The latter includes the implementation of the enumeration algorithm.