###----------------------------------------------------------------###
## The Gaussian white noise example from P1_fig_05.
## Load the required library
library(localgaussSpec)
###----------------------------------------------------------------###
## Specify the directory where the file-hierarchy will be stored.
main_dir <- c("~", "LG_DATA_scripts", "P1_fig_05")
## Note that 'main_dir' only contains the specification of the
## in-hierarchy part of the required path, and this path is stored as
## a vector. The reason for this is that it should be possible to
## move the storage directory 'main_dir' to another location on your
## computer, or even move it to a computer using another OS than the
## one used for the original computation.
###----------------------------------------------------------------###
## Simulate the data to be investigated.
## Simulate 'nr_samples' samples of length 'N' from the time series
## corresponding to 'TS_key', and save it into the file-hierarchy.
nr_samples <- 100
N <- 1974
TS_key <- "rnorm"
.seed_for_sample <- 124
set.seed(.seed_for_sample)
## Generate the sample.
.TS_sample <- TS_sample(
TS_key = TS_key,
N = N,
nr_samples = nr_samples,
.seed = NULL)
rm(nr_samples, N, .seed_for_sample)
## Create a unique 'save_dir' and save 'TS_sample' to the
## file-hierarchy.
save_dir <- paste(TS_key,
digest::digest(.TS_sample$TS),
sep = "_")
## Save to file and update file-hierarchy.
.TS_LG_object <- TS_LG_object(
TS_data = .TS_sample,
main_dir = main_dir,
save_dir = save_dir,
.remove_ties = TRUE)
rm(TS_key, .TS_sample, save_dir, main_dir)
###----------------------------------------------------------------###
## Compute the local Gaussian autocorrelations.
## This requires a specification of the desired points, the bandwidth
## and the number of lags. WARNING: The type of approximation must
## also be specified, i.e. the argument 'LG_type', where the options
## are "par_five" and "par_one". The "five" and "one" refers to the
## number of free parameters used in the approximating bivariate
## local Gaussian density. The results should be equally good for
## Gaussian time series, but the "par_one" option will in general
## produce dubious/useless results. Only use "par_one" if it is of
## interest to compare the result with "par_five", otherwise avoid it
## as it most likely will be a waste of computational resources.
.LG_type <- "par_five"
.LG_points <- LG_select_points(
.P1 = c(0.1, 0.1),
.P2 = c(0.9, 0.9),
.shape = c(3, 3))
lag_max <- 20
.b <- c(0.5, 0.75, 1)
## Do the main computation.
LG_AS <- LG_approx_scribe(
main_dir = .TS_LG_object$TS_info$main_dir,
data_dir = .TS_LG_object$TS_info$save_dir,
TS = .TS_LG_object$TS_info$TS,
lag_max = lag_max,
LG_points = .LG_points,
.bws_fixed = .b,
.bws_fixed_only = TRUE,
LG_type = .LG_type)
rm(.TS_LG_object, lag_max, .LG_points, .b, .LG_type)
###----------------------------------------------------------------###
## Send code to terminal that can be used to start the interactive
## inspection based on the shiny-application 'LG_shiny'. It might be
## of interest to save this to a file so an inspection later on does
## not require this script. Note that there are some tests in the
## code that try to prevent things that have already been computed
## from being computed once more, but the initial computation of
## '.TS_sample' will be performed every time this script is used.
LG_shiny_writeLines(
main_dir = LG_AS$main_dir,
data_dir = LG_AS$data_dir)
## Start the shiny application for an interactive inspection of the
## result. The use of 'shiny::runApp' is needed in order to start
## the shiny-application when this script is sourced.
shiny::runApp(LG_shiny(
main_dir = LG_AS$main_dir,
data_dir = LG_AS$data_dir))
###----------------------------------------------------------------###
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.