Features ideas and plans (i.e., TODO list) about report

[strengejacke]

Might be of interest: https://crsh.github.io/papaja_man/

[DominiqueMakowski]

I've been following their development for quite a while, but the scope seems quite different for now (but future bridges between the two could make sense, as they are positioned at an even higher level (document formatting) than report)

[DominiqueMakowski]

Came across this small but promising package.

They have some interesting ideas and made me think of some possible improvements:

Table legends

We could add a legend for report tables (consisting in the header chunk of the textual reports about model description), stored in the attributes, that we would display upon print.

Sample size info

We need to find a way to nicely add information about the sample size, flexible enough to deal with random groups

Factor levels and interactions formatting

We could better format factor levels and interactions:

• For factor levels, the idea would be to 1) detect whether a parameter is a factor level or a variable. 2) If a factor level, find the parent variable. 3) Display Something like:

  - Sepal.Width has an effect of ...
  - For Species:
     - versicolor has an effect of ...
     - setosa has an effect of

Instead of

  - Sepal.Width has an effect of ...
  - versicolor has an effect of ...
  - setosa has an effect of

However, the above proposition becomes trickier when involved in interactions.

• For interactions, instead of having Sepal.Width:versicolor, print either something like:

- The interaction between Sepal.Width and versicolor has an effect of ...

or

- For interactions:
  - Sepal.Width:versicolor has an effect of ...

But how to deal with interactions of interactions (e.g. Sepal.Length:Petal.Width:versicolor)?

[GegznaV]

Please, add functionality to report Cramer's V (and possibly other association measures for categorical variables). Here are some ideas on how Cramer's V could be reported:
http://rcompanion.org/handbook/H_10.html

[DominiqueMakowski]

@GegznaV It's noted, although we can't promise this feature anytime soon :)

@strengejacke As it is a very small function (see here), we could potentially include it in performance? (DescTools implements it with CIs)

[strengejacke]

There's a function in sjstats, xtab_statistics(), that calculates various statistics for contigency tables. There are also two functions that compute Cramer's V and Phi coefficient. CIs for Cramer are probably based on bootstrapping? Must look into DescTools.

I don't see Cramer's V as a function for performance, but rather parameters. Or another package that computes "various statistics", like Cronbach's Alpha etc?

[DominiqueMakowski]

Right, parameters seems more appropriate (until/if we create a better place for it).

[DominiqueMakowski]

Closing this for now, so we can start fresh