1.5. Philosophy and Design Goals of DYNAMIS-POP

DYNAMIS-POP is designed as a customizable portable platform. The model is intended to be used in many country contexts and care was given to make portability as easy as possible. This goal was reached by making most code modules as well as analysis scripts generic requiring code changes only in very few files. At the same time, DYNAMIS-POP is seen as a platform, which can be adapted, extended and refined by developers.

DYNAMIS-POP is based on data available for most countries. The typical data need are the microdata of a population census (or a sample thereof) and of a household survey like UNICEF’s Multiple Indicators Cluster Surveys (MICS) or Demographic and Health Surveys (DHS).

When implementing DYNAMIS-POP, we start from the ‚known‘, i.e. a micro-simulation implementation of the widely used cohort-component population projection model Demproj. Familiar parameters like age-specific fertility rates or mortality tables (typically available online) of macro projections can be used directly in the microsimulation model leading to identical projection results. More refined models which go beyond the scope of macro approaches are then added allowing for more detailed projections. The refined demographic models can be aligned to the base results or run without alignment. When aligned, the microsimulation model produces the same aggregate projections than a given macro model, but creates more realistic individual life courses. For example, births are realistically distributed by individual characteristics like education, partnership status, time since last birth, and parity, while the total number of births is kept identical with a given macro projection.

DYNAMIS-POP is reproducible. All required analysis scripts for parameter generation as well as the model code are available online.

DYNAMIS-POP is built and fully documented step-by-step. This holds true for both data analysis and model implementation. Typically, each step adds a new module and corresponds to an analysis script for estimating/generating the parameters of the added module. Each step is documented, allowing to use the documentation as a text book for microsimulation model development.

The modules of DYNAMIS-POP are kept as simple, transparent, and user friendly as possible, selecting models that allow intuitive parameters. For example, we use probabilities, hazard rates, and odds ratios whenever possible, avoiding complex regression coefficients.

DYNAMIS-POP is highly modular. Modules added at each step are typically self-contained and do not require changes in other modules. Also, various modules are optional and can be removed if not used. This allows model developers adding own modules and/or to replace existing modules by alternative implementations of modeled behaviors.

DYNAMIS-POP is freely available. This includes all required software tools and packages for model developers, like the Modgen programming language developed and maintained at Statistics Canada, the community version of Microsoft Visual Studio, the DYNAMIS model code and executable, as well as the statistical analysis package R. Downloads are also provided for data files for an imaginary country which can be used for training and illustration.

DYNAMIS-POP has a Graphical User Interface (GUI) which allows easy scenario creation, editing of all model parameters, running scenarios, and display results. The GUI is fully documented by an integrated help system.

DYNAMS-POP produces rich model output. This includes an extensive set of ready-made tables displayed within the user interface. Table outputs can also be exported to Excel format for post-processing and graphical representation. Also, the model produces micro-data output for post-processing by statistical packages like R. Another output type is a tracking database used to visualize individual life courses using the BioBrowser tool.

DYNAMIS-POP together with its documentation can serve as a text-book toolbox for microsimulation modeling and training. The model covers a wide a variety of approaches typically used in micro-simulation, including hazard regression, logistic regression, life tables, and parametric models.