Table Set of Correlations (for manuscripts)

[wibeasley]

(Spun off from Issue #97) After today's meeting with @andkov & @ampiccinin,

After considering it some more, @ampiccinin likes the existing plan plan for the tables containing correlations.

1. There are currently three versions of the same correlation information
   • the "study-specific manuscripts" version groups all the models of a study together, which will allow particiapnts to copy (a subset of) our tables into their manuscripts. To allow their readers to dive into their results, the cells will have the estimate, se, & p-value
   • the "CAR" version groups together all the process-pairs. To reduce clutter, the cells have only the estimates & significance-asterisks.
   • the dynamic & exhaustive version that will allow interested parties to make any comparison they want. At least for a few years, the Javascript table will be hosted on an IALSA website and on GitHub. At least for a few decades (theoretically), the whole repo will archived by the third-party Zenodo. (Here's an example of what we did for a recent textbook.)
2. The gray columns won't be included in the final tables. Because they're constant values, it can be factored out into the table title.
3. The xlsx file we sketched can be downloaded after clicking the 'View Raw' button.

[andkov]

consider computing the se during Fisher transform, so that the shape of the COMPUTED correlation conforms to the est/se/wald/pval format provided by Mplus:

[andkov]

[andkov]

TODO: research how MPLus calculates standard errors for correlations. Does it use covariances?

[andkov]

http://stackoverflow.com/questions/39307289/tableau-like-grouped-table-in-r-for-markdown