Results for categorical levers are indices not the actual values

[wxyeah]

The optimization results for categorical levers are indices not the actual values. To avoid confusion, I suggest either fixing this to return the actual values or clarifying this in the documentation.

I'm using version 3.0.0. A MWE to reproduce the issue is given below. The known optimum solution should be x = 30 (index = 2) and y = 2.0 (index = 2).

`
from ema_workbench import (
Model,
CategoricalParameter,
ScalarOutcome,
SequentialEvaluator,
Sample,
RealParameter
)

Model

def toy_model(x, y, b):

objective = -(x + y) + b

# return objective
return {"objective": objective}

EMA model

model = Model("Toy", function=toy_model)

model.uncertainties = [
RealParameter("b", 0.0001, 0.0005),
]

reference_SoW = Sample('reference',
b = 0.0003,
)

model.levers = [

CategoricalParameter(
    "x",
    (10, 20, 30)
),

CategoricalParameter(
    "y",
    (0.5, 1.0, 2.0)
)

]

model.outcomes = [

ScalarOutcome(
    "objective",
    ScalarOutcome.MINIMIZE
)

]

Optimization

with SequentialEvaluator(model) as evaluator:

results_toy = evaluator.optimize(

    nfe=20,
    searchover="levers",
    epsilons=[0.01],
    directory="toy_archive",
    filename="archive_test_toy.tar.gz",
    reference = reference_SoW

)

`

[quaquel]

I agree that this is something that needs to be addressed. My hunch is that I take the raw data from the experiments, rather than running them through their translation method. Should not be too difficult to address.