Bug 130835 - Adding multivariate correlation analysis for nominal and parametric variables
Summary: Adding multivariate correlation analysis for nominal and parametric variables
Status: NEW
Alias: None
Product: LibreOffice
Classification: Unclassified
Component: Calc (show other bugs)
(earliest affected)
Hardware: All All
: medium enhancement
Assignee: Not Assigned
Depends on:
Blocks: Data-Statistics
  Show dependency treegraph
Reported: 2020-02-21 12:55 UTC by Foad S. Farimani
Modified: 2022-10-23 19:46 UTC (History)
4 users (show)

See Also:
Crash report or crash signature:


Note You need to log in before you can comment on or make changes to this bug.
Description Foad S. Farimani 2020-02-21 12:55:44 UTC
Following this Tweet thread:


I was asked 


to open a feature request here.

Basically, the issue is that for multivariate correlation analysis at the moment Andrew F. Hayes 's SPSS macro, PROCESS, is the de facto tool used in the practice. As truly mentioned in this paper by Carden et al:


"There do exist a handful of ready-made solutions for implementing [Johnson-Neyman Technique] JN, but they suffer from one of several drawbacks. First, the highest-quality existing implementations are for software solutions requiring expensive licenses. The best known is the PROCESS add-on (Hayes, 2013b) for SAS and SPSS. There are a few problems with this. First, some researchers do not have access to SAS or SPSS, which are prohibitively expensive. ... To be sure, there are free solutions, but they suffer from at least one of the following drawbacks: the graphics are of bad quality, the process is not streamlined in that it requires the user to complete the linear regression and creation of the JN figure separately, or the user experience is intimidating to those without programming experience. For example, the probemod package (Tan, 2015) for R produces a graphic that would require considerable editing before being suitable for publication, and the linear regression must be completed separately beforehand, so the process is not as streamlined as it could be. The “rockchalk” package (Johnson, 2016) produces a higher-quality graphic but still requires the linear regression to be completed beforehand. Last but not least, Kristopher Preacher maintains the website www.quantpsy.org (Preacher et al., 2006) that allows one to run the JN technique. However, the user must manually enter the coefficients, coefficient variances, and so forth. As these must come from a software program, such as SPSS, one must toggle between software and website. Thus, it is not as seamless as it could be, and it may be prone to error when users enter their coefficients. Additionally, the site produces figures that are not publication-ready."

the situation for multivariate correlation analysis is pretty cumbersome. However, Condal et al., in my humble opinion, cease to address the concerns and surprisingly try to replace one proprietary software, SPSS, with another, Microsoft Excel:

"CAHOST, our workbook for implementing JN, is freely available, familiar to researchers of all levels, automates the linear regression, and produces publication-ready graphics."

Similarly, many people don't have access to the expensive MS products or, like me detest using any proprietary software at all. So here my proposal is to reimplement Hayes's PROCESS macro for LibreOffice Calc. Some of the Free, Libre, And Open Sorce (FLOSS) reimplementations of the methodology are:

1. Quentin André Python library PyProcessMacro https://github.com/QuentinAndre/pyprocessmacro
2. R package processR by Keon-Woong Moon: https://github.com/cardiomoon/processR
3. processr R package by Mark White: https://github.com/markhwhiteii/processr/

This I believe would significantly increase LibreOffocice's user base, especially in the humanities who are, stereotypically, more convenient with GUI programs rather than programing languages.
Comment 1 Foad S. Farimani 2020-02-23 16:26:44 UTC
I posted this question here


to start understanding the math behind the JN method
Comment 2 Xisco Faulí 2020-03-17 10:24:01 UTC
Moving to NEW