skipTrack (version 0.1.2)

skipTrack.results: Get tables of Inference results from skipTrack.fit

Description

This function calculates inference results on Betas, Gammas, and cijs based on the provided MCMC results. It returns summaries such as credible intervals for Betas, Gammas, wald-type confidence intervals for cijs, and Gelman-Rubin PSRF diagnostics for all 3. Note that true values and converage are included in the output if trueVals is supplied, but otherwise not.

Usage

skipTrack.results(stFit, trueVals = NULL, burnIn = 750)

Value

A list containing the following elements:

Betas

data.frame with 95% credible intervals and (if trueVals is supplied) true values for Betas and Coverage tag.

Gammas

data.frame with 95% credible intervals and (if trueVals is supplied) true values for Gammas and Coverage tag.

cijs

data.frame with Wald-type 95% confidence intervals and (if trueVals is supplied) true values for cijs and Coverage tags.

Diagnostics

data.frame with ess and gelman-rubin diagnostics from genMCMCDiag package, for parameter sets 'Betas', 'Gammas' and 'cijs'.

Arguments

stFit

Object result of skipTrack.fit function.

trueVals

Optional named list containing true values for Betas, Gammas, and cijs. (Also can use output of skipTrack.simulate)

burnIn

Number of MCMC iterations to discard as burn-in per chain.

Examples

#Simulated data
simDat <- skipTrack.simulate(n = 100, skipProb = c(.7, .2, .1))

#Run model fit (should typically run with much more than 50 reps)
modFit <- skipTrack.fit(Y = simDat$Y, cluster = simDat$cluster, chains = 2, reps = 50)
modFit

# If using simulated data (which includes access to ground truth):
#
skipTrack.results(modFit, trueVals = simDat, burnIn = 25)
#Recommended burnIn with real data is at least 750
#
# If not using simulated data:
#
skipTrack.results(modFit, burnIn = 25)
#Recommended burnIn with real data is at least 750