Maintain probability mass for continuous values as if they were another nominal value.
For example it we might have
P(a="v1") = 0.5
P(a=Continuous value) = 0.5
Where the continuous values maintain a normal distribution so the probability of a specific continuous value can be computed. In the case where there are not nominal values, this will be identical to the current implementation.
However, this should allow for mixing of nominal and numeric values. This will be particularly important for maintaining a probability of None, but I could imagine it might come in useful in other situations as well.
Maintain probability mass for continuous values as if they were another nominal value.
For example it we might have
P(a="v1") = 0.5 P(a=Continuous value) = 0.5
Where the continuous values maintain a normal distribution so the probability of a specific continuous value can be computed. In the case where there are not nominal values, this will be identical to the current implementation.
However, this should allow for mixing of nominal and numeric values. This will be particularly important for maintaining a probability of None, but I could imagine it might come in useful in other situations as well.