Alignment functions and other utilities for data.table time-series.

Motivation

data.table allows the creation of time-series by creating a first column with a temporal type. The functions in this package assume this temporal type is nanotime.

All the standard data.table functions can be used with such a time-series. The goal of this package is to consolidate this functionality with functions that are specialized for time-series such as alignment functions.

The alignment functions that are proposed in this package are particularly versatile and allow to build higher-level functions quite easily. And example of this is shown in the package itself with the function frequency.

Examples

The following function creates a time-series of 100 rows spaced every 1 seconds. It then aligns this time series on a time vector that is spaced every 10 seconds and applies the column means to the observations that are being aligned.

library(dtts.utils); library(data.table)
one_second  <- 1e9
cols <- 3
rows <- 100
t1 <- nanotime(1:rows * one_second)
dt1 <- data.table(index=t1, matrix(1:(rows*cols), rows, cols))
t2 <- nanotime(1:10 * one_second * 10)
align(dt1, t2, start=-10*one_second, func=function(x) colMeans(as.data.frame(x)))

Status

The package is in the very early stages and is largely untested.

See the issue tickets for an up to date list of potentially desirable, possibly planned, or at least discussed items.

Installation

remotes::install_github("lsilvest/dtts.utils")

Author

Leonardo Silvestri

License

GPL (>= 2)