Accession Number:



Statistical Validation of a New Python-based Military Workforce Simulation Model

Descriptive Note:

Technical Report

Corporate Author:

Defence Research and Development Canada Ottawa Canada

Report Date:


Pagination or Media Count:



The Canadian Department of National Defence uses military workforce modelling and simulation to inform the decision-making process with regard to the management of military personnel. Defence Research and Development Canada DRDC has a suite of workforce models that have been used in a variety of studies over many years to answer questions, generate forecasts and analyse various scenarios under consideration. Over time, these simulation models have been refined, errors corrected, and results validated against historical data. Consequently, the level of confidence in these models is very high. As computer technology advances, the need arises periodically to update these models to take advantage of newer technology and to provide more advanced capabilities. DRDC is updating its workforce modelling and simulation technology through the development of a Python-based discrete event simulation environment that is intended to replace various commercial simulation software products in which existing DRDC workforce models have been built. DRDC has begun the process of rebuilding select workforce models in this new environment. Rebuilding a simulation model using new technology usually results in some loss of confidence in the model even if the features of the new technology are impressive. This is due to the possibility of reintroducing errors, inexperience with the new technology, and a lack of validation against hard data. To establish confidence in the new model implementation, it is necessary to validate its equivalence to the older trusted version. This can be done by subjecting both implementations to identical input scenarios and comparing simulation output. However, because workforce models frequently make use of random effects modelled using probability distributions e.g. age at recruitment, release events, and course failures, variability in simulation output from the same input is expected.

Subject Categories:

Distribution Statement: