Rev Language Reference

TraceTree - Trace of trees.

Stores a tree trace, usually produced by passing a variable of type `Tree` to the `mnFile` monitor in an MCMC or MCMCMC run.

Details

Individual trees can be accessed as follows: - `getTree(n)`: returns the `n`th tree. - `getTree(n, post=TRUE)`: returns the `n`th post-burnin tree. - `getTrees()`: returns the entire vector of post-burnin trees. The methods `.computeEntropy()`, `.computePairwiseRFDistances`, `.getUniqueTrees()`, `.isTreeCovered()`, and `.summarize()` first construct a credible set of trees, and then perform the corresponding operation on this set. A credible set is constructed by ordering all trees in the trace by their posterior probability, and adding them to the set one by one until the included trees accumulate the desired probability specified by `credibleTreeSetSize`. By default, this argument is set to 0.95; to perform a given operation on all trees in the trace, set it to 1. The credible set methods also take a `probabilistic` argument that determines how the credible set should be constructed when it is not possible to obtain a cumulative probability precisely equal to `credibleTreeSetSize`. This will be the case when the probability p of the first n trees falls short of it, and the probability (p + q) of the first n + 1 trees exceeds it (Huelsenbeck & Rannala 2004). If `probabilistic=TRUE` (default), the (n + 1)-th tree will be included in the set with probability (`credibleTreeSetSize` - p) / q. It will always be included if `probabilistic=FALSE`. Note that when `probabilistic=TRUE`, the methods in question may yield different results under different random seeds, and they may even construct an empty credible set a certain fraction of the time if the trace is dominated by a particular tree whose probability exceeds `credibleTreeSetSize`. An exception is thrown when an empty credible set is encountered by a method that does not expect one.

Example

# read a tree trace and ignore the first 10 samples
thinned_trees = readTreeTrace("my_filename.tree", offset = 10, thinning = 10, burnin = 0.5)

thinned_trees.getTree(1) # Returns the 11th tree (offset + 1) in the file
thinned_trees.getTree(2) # Returns the 21st tree (offset + 1 + thinning) in the file

thinned_trees.getTrees()[1] # Returns the first sampled tree after excluding the burnin fraction

Methods

See Also

Edit this page