itsrainingdata/sparsebnutils

Utilities for learning sparse Bayesian networks

sparsebnutils's Issues

Improve reIndex methods

Some of the methods currently assume input is given either via R-style or C-style indexing. It would be better to avoid this assumption and check this carefully for each input.

Example 1

select.parameter compatibility for discrete output

Improved compatibility with Matrix objects

The following internal methods are not supported for Matrix input, but should be:

• as.sparse
• as.edgeList

Improve output of show.parents with long node names

Long node names are truncated to 4 chars by default in show.parents. Add an extra argument to allow this to be adjusted by the user.

sparsebnUtils/R/sparsebnUtils-show.R

Line 32 in 9f84eb7

edges.str <- edgeList_to_node_names(x, 4) # convert edgeList to use node names

Custom RNG for random.dag is broken

Error when passing a custom function for generating random numbers to random.dag:

> random.dag(5,5, call = runif)
Error in call() : argument "n" is missing, with no default
Called from: call()

Inference

The description of fit_multinom_dag() says that given the structure of a Bayesian network it estimates the parameters using multinomial logistic regression. Is this the function used to find inference? Also, what exactly does this function output?

Add call component to sparsebnFit

Similar to default R modeling functions such as lm, glm. See ?match.call.