biips_smc_sensitivity: Sensitivity analysis of parameters with SMC

In biips/rbiips: Bayesian Inference with Interacting Particle Systems

Description

Sensitivity analysis of parameters with SMC

Usage

biips_smc_sensitivity(object, param_values, n_part, ...)

Arguments

object	biips model object as returned by biips_model.
param_values	named list of numeric values. Each member is named after a parameter for which we want to study sensitivity. Each member value is a vector or array whose last dimension is equal to the number of values for which we want to study sensitivity.
n_part	integer. Number of particles.
...	Additional arguments to be passed to the SMC algorithm such as rs_thres and rs_type. See biips_smc_samples for more details.

Value

biips_smc_sensitivity returns a list with the following members:

log_marg_like	vector of log marginal likelihood estimates at the different values of the parameters.
log_marg_like_pen	vector of penalized log marginal likelihood estimates at the different values of the parameters.
max_param	parameters value with maximum log_marg_like.
max_log_marg_like	maximum log_marg_like value.
max_param_pen	parameters value with maximum log_marg_like_pen.
max_log_marg_like_pen	maximum log_marg_like_pen value.

See Also

biips_model, biips_smc_samples

Examples

modelfile <- system.file('extdata', 'hmm.bug', package = 'rbiips')
stopifnot(nchar(modelfile) > 0)
cat(readLines(modelfile), sep = '\n')

data <- list(tmax = 10, p = c(.5, .5), logtau_true = log(1), logtau = log(1))
model <- biips_model(modelfile, data)

n_part <- 50
logtau_val <- -10:10
out_sens <- biips_smc_sensitivity(model, list(logtau = logtau_val), n_part)

par(mfrow = c(2, 1))
plot(logtau_val, out_sens$log_marg_like, type = 'l', col = 'blue',
     xlab = 'logtau', ylab = 'log p(y|logtau) ', main = 'SMC sensitivity')
points(data$logtau, min(out_sens$log_marg_like), pch = 17, col = 'green')

plot(logtau_val, out_sens$log_marg_like_pen, type = 'l', col = 'blue',
     xlab = 'logtau', ylab = 'log p(y|logtau) + log p(logtau)')
plml <- out_sens$log_marg_like_pen
ymin <- min(plml[is.finite(plml)])
points(data$logtau, ymin, pch = 17, col = 'green')

biips/rbiips documentation built on Nov. 28, 2020, 2:12 p.m.