simNodes: Basic nimbleFunctions for calculate, simulate, and getLogProb...

In thirdwing/nimble: Flexible BUGS-compatible system for hierarchical statistical modeling and algorithm development

Description

simulate, calculate, or get existing log probabilities for the current values in a NIMBLE model

Usage

simNodes(model, nodes)

Arguments

model	A NIMBLE model
nodes	A set of nodes. If none are provided, default is all model$getNodeNames()

Details

These are basic nimbleFunctions that take a model and set of nodes and return a function that will call calculate, simulate, or getLogProb on those nodes. Each is equivalent to a direct call from R, but in nimbleFunction form they can be be compiled and can be put into a nimbleFunctionList. For example, myCalc <- calcNodes(model, nodes); ans <- myCalc() is equivalent to ans <- calculate(model, nodes), but one can also do CmyCalc <- compileNimble(myCalc) to get a faster version.

In nimbleFunctions, for only one set of nodes, it is equivalent or slightly better to simply use calculate(model, nodes) in the run-time code. The compiler will process the model-nodes combination in the same way as would occur by creating a specialized calcNodes in the setup code. However, if there are multiple sets of nodes, one can do the following:

Setup code: myCalcs <- nimbleFunctionList(calcNodes); myCalcs[[1]] <- calcNodes(model, nodes[[1]]); myCalcs[[2]] <- calcNodes[[2]]

Run code: for(i in seq_along(myCalcs)) {ans[i] <- myCalcs[[i]]()}

Author(s)

Perry de Valpine

thirdwing/nimble documentation built on May 31, 2019, 10:41 a.m.