Skip to contents

mtsdesc

The mtsdesc R package is being developed to provide mtsdesc provides commonly needed utilities when working with a set of time series, i.e. multi-panel plots and multi-column summary- and testing-tables. The package also has methods to work with residuals of fitter values of multivariate time series models.

Disclaimer: this package is in the early most stage of life. It hasn’t been thoroughly tested and can present several bugs. I don’t recommend using it for large-scale projects, yet.

Please report any problems as a GitHub issue. Other comments can be posted as a GitHub discussion, or sent in my email below. Thank you!

Author: Ricardo Semião e Castro ().

Introduction

The main package functions are either plotting ones ggmts_*() or summary-table ones tbmts_*(). The workflow of both is similar:

  • They receive either a set of time series or a multivariate time series model.
  • Then, the data is extracted. For the former, the series, for the latter, residuals, fitted, or predicted values. This is done with the internal generic functions *_setup().
  • Then, the data is processed into a plot or table. The output can be customized with args_* arguments, that inject options to each dimension of the plot or table. The result can be further processed by hand after the function call.

As of now, the only format of time series accepted is a data.frame-ish object, and for model, only varest from the vars package. But, the separation of the data extraction in *_setup() and the output generation means that you can provide your custom methods for *_setup() (eg. *_setup.mts()) and the rest of the function should work. I intent to do a tutorial guide on how to do these extensions in the future.

Additionally, the package was originally focused on VAR models, such that the plotting functions are still called ggmts_*(), and there is not yet table functions tbmts_*().

Installation

You can install the development version of mtsdesc from GitHub with:

# install.packages("devtools")
devtools::install_github("ricardo-semiao/mtsdesc")

Development Comments

The package has some nice features, with already some testing and documentation. But, I’ve recently changed its purpose to be more general on mts, and adopted new testing conventions. Thus, it will go through a major refactor in the next months.

Some of the more important features I plan to add in the future are:

  • Major refactor:
    • Start using rlang and glue APIs wherever possible, especially in the “setup” suite.
    • Add tidyselect capabilities.
    • Choose a better convention for faceting variable names and “serie(s)”, “var(s)” or “col(s)” namings.
    • Add vars to suggests. Add check_installed for suggests.
    • Have better default methods for *_setup() functions (assuming dataframes).
    • Rethink the test functions list scheme.
    • Add *_setup() documentation in the function docs, talking about existing methods and directing to their engines (“see: ?vars:::predict”).
    • Define a convention for layers orders.
    • Study importing whole rlang and ggplot2.
  • New features:
    • Create the table functions tbmts_*() for summary and testing tables.
    • Add option to remove facets in graphics with a single serie.
    • Study which plots could have different versions like _colored.
  • More thorough tests:
    • Add tests for “helper” and “setup” functions.
    • Add tests for test function itself, and also for documentation roxy.
    • Specify what errors to expect in expect_error calls.
    • Directly test for elements of the ggplots returned by functions.
    • Study including warnings if arguments in args_* aren’t used by the relevant function.
  • More thorough documentation:
    • Annotate all conventions in a .md file.
    • More tips on how to customize results in the vignette and in the function docs.
    • Add better examples, add an example.qmd vignette.
    • Solve some less important lintr notes.

Note that this package: