[BUG] Does not work with models that have a different Bayesian Network Structure/Number of variables

[Wh0DKnee]

Description

The project only works for models with a specific Bayesian Network structure with seven parameters. For example, it does not work with the glider_v1p2.txt model, as there are only five variables in the model.

Reproducibility

Create an ini file for the input that uses the glider_v1p2.txt model, use that ini file in RUN_DAA_EncounterModelTool_serial.m as the "parameterFile".

A bunch of asserts will fire. When commenting out the asserts, an error will occur because the startDefault variable in line 48 in generateDAAEncounterSet.m has hard-coded dimensions (7), which doesn't match the glider model.

Expectation

Should work with any model.

Files

em-pairing-uncor-importancesampling\Encounter_Generation_Tool\generateDAAEncounterSet.m
em-pairing-uncor-importancesampling\RUN_DAAEncounterModelTool_serial.m
em-pairing-uncor-importancesampling\Encounter_Generation_Tool\checkEncounterModelInputs.m

Fix

I've created a fork with a hack as a fix (hardcoded so that it works with models that only have 5 random variables): Fork

[Wh0DKnee]

Update: I also believe that lines 85 and 131 in generateDAAEncounterSet.m do not work for other models. As I understand it, the indices in "idxZeroBoundaries" correspond to the hard-coded random variables of the model. I am not 100% sure though.

[aweinert-MIT]

idzZeroBoundaries specifies how model variables are sampled. Either the index of the variable bin (e.g. L = 1) is return or a value sampled within the desired bin (e.g. L = 200)