| [stem-dev] Discussion |
Hi all,
a few things came up on todays STEM call and I just wanted to summarize what was discussed. They all have to do with the relationship between disease models and population models and how they are composed in a STEM scenario.
1. When a disease model is initialized (creating and initializing the disease model label and label values needed to keep track of the dynamically changing state of the disease), it depends on having a population model (of the same population identifier) with its label/label values already initialized beforehand. If it cannot find one, it automatically generates an instance of a StandardPopulationModel with zero background birth/death rate. The question was whether this is perhaps more confusing to the user than helpful, and perhaps it's better to force the user to explicitly add the population model manually. This would also make it clear to the user what the background birth and death rate for the population is. An alterative suggestion was made to make it an option when creating a disease model whether to also automatically add the zero birth/death rate population model for it.
2. Right now, population models must be located beneath the the disease model in the model hierarchy. The only real reason why they can't be at the same level is that for Standard Population Models (with fixed background birth/death rate), the population numbers are adjusted to correspond to the starting date of the sequencer. So if the population data is valid 2006, and the sequencer starts in 2011, the numbers are advanced five years given the background birth/death rate. The disease model must be aware of this when it is initialized so it must be above the population model. Again, the question is whether this is confusing the users more than helping them. Also, since we have many other types of population models that does not change the initial population numbers from the start of the sequencer (e.g. seasonal, mosquito), perhaps we should get rid of this feature altogether.
3. If we were to get rid of the feature, perhaps replacing it with an alternative method such as using a "population re-scaler" is a better option. This has the benefit of working for any type of population model used.
Regards,
/ Stefan
Stefan Edlund
Public Health and Computer Science Research
IBM Almaden Research Center
(408) 927-1766 edlund@xxxxxxxxxxxxxxx