stagedtrees (version 2.1.0)

compare_stages: Compare two staged event tree

Description

Compare two staged event trees, return the differences of the stages structure and plot the difference tree. Three different methods to compute the difference tree are available (see Details).

Usage

compare_stages(
  object1,
  object2,
  method = "naive",
  return_tree = FALSE,
  plot = FALSE,
  ...
)
hamming_stages(object1, object2, return_tree = FALSE)
diff_stages(object1, object2)

Arguments

object1

an object of class sevt.

object2

an object of class sevt.

method

character, method to compare staged event trees. One of: "naive", "hamming" or "stages".

return_tree

logical, if TRUE the difference tree is returned.

plot

logical.

...

additional parameters to be passed to plot.sevt.

Value

compare_stages: if return_tree = FALSE, logical: TRUE if the two models are exactly equal, otherwise FALSE. Else if return_tree = TRUE, the differences between the two trees, according to the selected method.

hamming_stages: if return_tree = FALSE, integer, the minimum number of situations where the stage should be changed to obtain the same models. If return_tree = TRUE a stages-like structure showing which situations should be modified to obtain the same models.

diff_stages: a stages-like structure marking the situations belonging to stages which are not the exactly equal.

Details

compare_stages tests if the stage structure of two sevt objects is the same. Three methods are available:

• naive first applies stndnaming to both objects and then simply compares the resulting stage names.

• hamming uses the hamming_stages function that finds a minimal subset of nodes which stages must be changed to obtain the same structure.

• stages uses the diff_stages function that compares stages to check whether the same stage structure is present in both models.

Setting return_tree = TRUE will return the stages difference obtained with the selected method. The stages difference is a list of numerical vectors with same lengths and structure as stages(object1) or stages(object2), where values are 1 if the corresponding node has different (with respect to the selected method) associated stage, and 0 otherwise.

With plot = TRUE the plot of the difference tree is displayed.

If return_tree = FALSE and plot = FALSE the logical output is the same for the three methods and thus the naive method should be used since it is computationally faster.

hamming_stages finds a minimal set of nodes for which the associated stages should be changed to obtain equivalent structures. To do that, a maximum-weight bipartite matching problem between the stages of the two staged trees is solved using the Hungarian method implemented in the solve_LSAP function of the clue package. hamming_stages requires the package clue.

Examples

# NOT RUN {
data("Asym")
mod1 <- stages_bhc(full(Asym, lambda = 1))
mod2 <- stages_fbhc(full(Asym, lambda = 1))
compare_stages(mod1, mod2)

##########
m0 <- full(PhDArticles[, 1:4], lambda = 0)
m1 <- stages_bhc(m0)
m2 <- stages_bj(m0, distance = "totvar", thr = 0.25)
diff_stages(m1, m2)
# }