poststat {poolABC}

Calculate point estimates from the posterior distribution

Description

Given a set of samples from the posterior distribution, computes the mean, median and mode of the posterior.

Usage

poststat(posterior, limits, method, wtreg = NULL)

Arguments

posterior

is a matrix or a vector with samples from the posterior distribution obtained from ABC parameter estimation. If this input is a matrix, then each row should correspond to an accepted simulation (size S) and each column to a different parameter.

limits

is a vector if there is only one parameter or a matrix if there are multiple parameters. In this latter instance, each row should correspond to a different parameter. In either instance, and considering matrix rows as vectors, then the first entry of the vector should be the minimum value of the prior and the second entry should be the maximum value (for any given parameter).

method

either "rejection" or "regression" indicating whether a rejection sampling algorithm or a local linear regression algorithm were used during ABC parameter estimation.

wtreg

is a required numeric vector if the method is "regression". It should contain the weights for each accepted simulation (size S).

Details

If method is "regression", the regression weights must also be made available. These will be used to compute the weighted mean, weighted median and weighted mode of the posterior.

Value

a matrix with the mode, median and mean of the posterior distribution for each parameter. Each point estimate is a different row and each parameter a different column.

Examples

# load the matrix with parameter values
data(params)
# load the matrix with simulated parameter values
data(sumstats)
# load the matrix with the prior limits
data(limits)

# select a random simulation to act as target just to test the function
target <- sumstats[10 ,]
# we should remove the random simulation from the sumstats and params matrices
sumstats <- sumstats[-10, ]; params <- params[-10, ]

# parameter estimation for a single target
myabc <- singleABC(target = target, params = params, sumstats = sumstats,
limits = limits, tol = 0.01, method = "regression")

# compute point estimates from the posterior distribution
poststat(posterior = myabc$adjusted, limits = limits, method = "regression", wtreg = myabc$wt)