plot_individual_changepoint_posterior {BCT} | R Documentation |
This function plots the conditional posterior distribution of the change-points locations given a specific number of change-points.
plot_individual_changepoint_posterior(res, burn, pm, l = NULL)
res |
the output obtained from the Metropolis-Hastings algorithms (either from infer_fixed_changepoints or infer_unknown_changepoints). |
burn |
the proportion of the samples discarded as burn-in. |
pm |
the desired range around the MAP location for each change-point location. |
l |
condition on the number of change-points. If not initialised, the function expects as input the results obtained from the infer_fixed_changepoints function. |
plots of the empirical posterior distributions of the change-points given a specific number of change-points.
infer_fixed_changepoints
, infer_unknown_changepoints
# Use as an example the el_nino dataset.
# Run the function with l_max = 3 change-points, a maximum depth of 5 and the [0, 1] alphabet.
# The sampler is run for 10000 iterations.
res_unknown <- infer_unknown_changepoints(el_nino, 3, 5, c("01"), 100, fileName = NULL)
# Because l_max = 3 , there can be 0, 1, 2 or 3 changes.
# Let's see the posterior distribution on the number of changes
plot_changepoint_posterior(res_unknown, 0.2)
# The MAP l is 2. Let's see the distribution of changes given l = 2.
plot_individual_changepoint_posterior(res_unknown, 0.2, 20, 2)
# One can also see the distribution of changes given l = 1.
plot_individual_changepoint_posterior(res_unknown, 0.2, 500, 1)
# This function can be also used with the infer_fixed_changepoints
# Assume l = 2.
res_fixed <- infer_fixed_changepoints(el_nino, 2, 5, c("01"), 100, fileName = NULL)
# The function is now called without l = 2 as the number of changes is fixed
# (all sampled vectors have 2 values).
plot_individual_changepoint_posterior(res_fixed, 0.2, 20)