In StevenGolovkine/simulater: Simulate functional data based on real datasets

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

In this vignette, we will show how to generate realistic datasets. For this example, the data will be simulated based on the powerconsumption dataset provides in the package.

# Load the packages
library(simulater)

First, we need to estimate the mean, covariance and noise of the real dataset. Pay attention that these functions may take some time to execute.

mu <- learn_mean(powerconsumption)

cov <- learn_covariance(powerconsumption)

noise <- learn_noise(powerconsumption)

Then, we can generate some curves using the previous estimated parameters.

X <- generate_data(10, 100, mu, cov, noise, exp(-5.5), NULL, 0.2, 1)

plot(X[[1]]$t, X[[1]]$x, type = 'l', ylim = c(230, 250))
for(i in 2:9){
  lines(X[[i]]$t, X[[i]]$x, col = i)
}

Now, we will (visually) compare true data with a simulated realization. The true data from the powerconsumption dataset is plotted in blue, while a generated curve is plotted in red.

X <- generate_data(1, 1440, mu, cov, noise, exp(-5.5), NULL, 0.2, 1)

plot(X[[1]]$t, X[[1]]$x, type = 'l', col = 'red', ylim = c(230, 250))
lines(seq(0, 1, length.out = 1440), matrix(powerconsumption[1,]), col = 'blue')

StevenGolovkine/simulater documentation built on April 4, 2022, 5:04 a.m.