rsa.tuneDepthAlpha: Tune depth and alpha hyperparamters

In benpeloquin7/rrrsa: Run Rational Speech Act Models of Pragmatic Inference

Description

Return a list with number of alpha * depths elements each element includes a tuple of (correlation, alpha, depth). Same basic call functionality as runDf()

Usage

rsa.tuneDepthAlpha(data, quantityVarName, semanticsVarName, itemVarName,
  groupName = NA, compareDataName, costsVarName = NA, priorsVarName = NA,
  depths = 1, alphas = 1, compareIndices = NA,
  usePriorEveryRecurse = TRUE)

Arguments

data,	tidied data
quantityVarName,	entity name we're quantifying over
semanticsVarName,	semantic values for inference computation
itemVarName,	unique items were comparing, probaby words
groupName,	grouping variable if we have one
compareDataName,	pragmatic judgments we're comparing to
costsVarName,	costs variable name
priorsVarName,	priors variable name
depths,	vector of depths (in integers) for tuning
alphas,	vector of alphas for tuning
compareIndices,	specific indices in data frame
usePriorEveryRecurse,	boolean incorporate priors during each full recursion

Value

list of length(alphas) * length(depths) tuples with (correlation, depth, alpha)

Examples

d <- peloquinFrank_5Alts
alphas <- seq(1, 3, by = 0.1)
depths <- 1:3
checkWords <- c("some", "all", "good", "excellent", "liked", "loved", "memorable", "unforgettable",
"palatable", "delicious")
compareIndices <- which(peloquinFrank_5Alts$words %in% checkWords)
results <- rsa.tuneDepthAlpha(data = d, groupName = "scale",
quantityVarName = "stars", itemVarName = "words",
semanticsVarName = "speaker.p", compareDataName = "e11",
compareIndices = compareIndices, alphas = alphas, depths = depths)
head(results)
best <- which.max(unlist(lapply(results, function(i) i[[1]][1])))
results[[best]]

benpeloquin7/rrrsa documentation built on May 12, 2019, 2:08 p.m.