ld-archer / E_FEM

This is the repository for the English version of the Future Elderly Model, originally developed at the Leonard D. Schaeffer Center for Health Policy and Microsimulation.
MIT License
3 stars 1 forks source link

Create barebones version of the FEM #61

Closed ld-archer closed 3 years ago

ld-archer commented 3 years ago

Cut down the current model to only include information on key variables.

To Keep (outputs and predictors):

Some other things are required (ID, weighting vars, some others I can't remember now).

On a new branch we will remove everything else from the model (leaving the .cpp files alone, it doesn't matter if we have defined variables that we are not including).

ld-archer commented 3 years ago

REMOVED:

Checking outputs now to see if the loss of variables have improved/worsened projections or any specific model. Paying particular attention to BMI as this was good before the change. If all is good, will move on to checking against external data using the AgeUK Almanac.

ld-archer commented 3 years ago

Speaking to Bryan last Monday (4th Jan 21), he reminded me there is an even easier way to run multiple simulations with different variable sets than what I have done already here. Instead of limiting the variables in the starting populations, we can limit what is actually transitioned and output by the model using the settings files instead.

Therefore, the 'barebones' branch with the reduced model has been replaced with a 'core' model on the multisim branch, about to be merged with development.

Some slight changes between the barebones and core model:

Running the core model is very similar to other configurations, with the command make core. Eventually, a new debug script (or adapted old one) will be written for the core models and runnable via make core_debug.