Decomposable model conditioning

Conditioning is easy only if the compositional model is decomposable. Then, it is enough to reorder it in a way that the conditional variable appears among arguments of the first probability distribution in the generating sequence and create a degenerated probability distribution over the variable with value value

conditioning(model, variable, value)

Arguments

model

Compositional model
variable

Name of discrete random variable
value

Value of discrete random variable

Value

change original model in its input

Examples

data(m)
conditioning(m, variable = "T", value = 1)
#> Error in conditioning.Model(m, variable = "T", value = 1): Model is not decomposable. Conditioning is not possible.
mDecomposable <- toDecomposable(m)

dTable(toDistribution(marginalize(mDecomposable, variables = c("W"), keep = TRUE, new = TRUE)))
#>    W MUDIM.frequency
#> 1: 2         0.62528
#> 2: 1         0.37472
conditioning(mDecomposable, variable = "T", value = 1)
marginalize(mDecomposable, variables = "W", keep = TRUE, new = FALSE)
dTable(toDistribution(mDecomposable))
#>    W MUDIM.frequency
#> 1: 2          0.7984
#> 2: 1          0.2016