Skip to content

Home

Jump to bottom
SteffenMoritz edited this page Jul 11, 2019 · 10 revisions

Welcome to the imputeTS wiki!

imputeTS is Open Source Software and we are happy about everyone who wants to contribute!

Our vision is to develop a software, which makes it easy to use advanced algorithms to replace missing values in time series.

This helps scientists and industrial users of different disciplines all around the world. Since probably nearly everyone working with data has experienced NA's already.

How to Contribute

If you are on this page, you probably have used imputeTS already. There are several ways to contribute:

  1. If you find bugs, try to open an Issue here (everyone that finds a bug will be mentioned in the package NEWS once it is fixed)

  2. Spread the knowledge (People can only use the package if they know about it - having tutorials and material online helps here )

  3. Contribute on GitHub (there are plenty of things that are planned for the future of this package)

About 3:

  • Directly make a pull request for small fixes and improvements (I'll check and directly merge them)

  • For larger plans it might make sense to discuss this first here. This is important to keep the package consistent and that nobody makes double or unneeded effort.

  • In general suggestions about new features / changes, ... are very welcome. If you directly can implement them it's even better.

About code quality:

  • Try to take the existing code as an example.

  • Try to use https://style.tidyverse.org/ as an orientation

  • I additionally use the R package styler to style my files after I programmed something

  • I additionally use the R package lintr to check the current file (but just as an orientation, no 100% conformity needed here)

  • If you perform changes or additions, adding some testthat cases makes usually sense

  • And finally the package of course still needs to pass devtools::check()

Future Development Plans

You can add whatever you think makes sense under suggestions to this list!

Rather soon:

  • Better plots using ggplot2 Moving the existing plots to ggplot2 would be nice. These just look better. But I am not sure what this would look like yet. Probably renaming them would then also make sense (since I anyway wanted to get away from the . syntax - plotNA.xxx). Could be plotNA_distribution, gg_distribution, ggNA_distribution, gg_nadistribution, plotna_distribution. Opinions and suggestions welcome... :-)

  • Better and more unit tests Might be interesting to add more variety to input data that is tested. I just realized, at the moment it is mostly positive values.

  • Additional vignettes

    • First I am thinking about Quick Start guide or "Getting started".
    • Then it might make sense to have something more imputation related - a vignette where it is explained how to choose a algorithm or which algorithm is a good choice for which problem - maybe also some real examples
    • Also some information on how different ts types (zoo, tsibble,... interact might be interesting) or to show how imputeTS can be used in data preparation workflows

  • Adding harmonic and geometric mean as option for na_mean (done)

  • Readme update Probably makes sense, to have at least an example where the pipe %>% operator is used also in the readme.

Rather later:

  • Additional algorithms (RNN, Pattern,..)
  • multiple imputation additions
  • confidence intervals
  • automatic algorithm selection

Suggestions:

Clone this wiki locally