Glossary¶
- Model Specification
- A yaml file that details all components and configuration necessary to run a particular model.
- Branch Configuration
- A yaml file that lists the count of input data draws and random seeds as
well as a set of simulation configuration options. When coupled with a
model specification, this file defines a set of different simulation
scenarios that can be run in parallel with the
psimulatecommand line utility. - Parameter Uncertainty
- Parameter uncertainty is uncertainty due to the input data. The Global Burden of Disease represents uncertainy distributions around the parameters it produces with draws. By running a simulation with several different draws of the input data, we can propagate the parameter uncertainty through our model and to our outputs.
- Stochastic Uncertainty
- Stochastic uncertainty is uncertainty due to the inherent variability in the model. This variability is represented in a variety of places by using random numbers to sample from distributions. Our simulations control stochastic uncertainty with random seeds. Stochastic uncertainty in simulations is deeply related to sampling uncertainty in a study. By holding the input parameters constant and varying the random seed, we can produce many realizations (or samples) of the same underlying population and use that to minimize the stochastic uncertainty.
- Input Draw
- A way of representing uncertainty in the input parameters. Rather than having an explicit distribution, each input draw of a parameter is a sample from the underlying distribution of that parameter in a population. Taken collectively, all the input draws form a numeric representation of the original parameter distribution.
- Random Seed
- A number used to seed the simulation’s random number generator. Two simulations run with the same random seed will produce the same random numbers for each decision made in the simulation on each time step for each person.
- Configuration Parameter
- A parameter of a given model specified in the
configurationblock of the model specification. Typically determined early on in the model development process, these parameters determine what scenarios can be run with a given model. Common configuration parameters include the target coverage of an intervention, what subset of a population the intervention targets, and how effective an intervention is.