Feature Request: Add contingency table statistics (e.g., odds ratio, risk ratio)

[Polkas]

What's the feature?

We would like to contribute new functionality to the {cardx} package for analyzing contingency tables, specifically to enable calculation and reporting of effect estimates such as the odds ratio (OR) and risk ratio (RR), along with confidence intervals.

Motivation

In clinical studies and similar settings, users often want quick calculation of association measures (e.g., OR, RR) between groups for binary/2x2 tables (such as adverse events vs. treatment).

Proposed features

New functions to produce contingency table statistics.

Request

• Would such a contribution be welcome?
• Do you have any design suggestions or guidelines I should follow?
• Are there open PRs or plans for this already, or is this a good place for me to contribute?

[ddsjoberg]

Thanks for the post @Polkas . Is there a package you prefer using for these calculations? If not, these may be good matches for an package with Roche-GSK-Lilly-Pfizer focussed on confidence intervals (https://github.com/GSK-Biostatistics/citools). cardx does not implement statistical methods, rather wraps them for ARDs.

[Polkas]

Thank you for your quick and helpful response.

I appreciate the pointer to citools and also wanted to mention [contingencytables](https://cran.r-project.org/web/packages/contingencytables/index.html), which is part of a published book and shows how far contingency table analysis can go.

Your feedback helped me understand the design philosophy and strategy behind cardx, especially the goal to reuse existing solutions and to avoid adding too many dependencies. This helps keep cardx lightweight and user-friendly, which is great. For reference, here is the current dependency tree:

> tools::package_dependencies("cardx", recursive = TRUE)[[1]]
 [1] "cards"      "cli"        "dplyr"      "glue"       "lifecycle" 
 [6] "rlang"      "tidyr"      "tidyselect" "utils"      "generics"  
[11] "magrittr"   "methods"    "pillar"     "R6"         "tibble"    
[16] "vctrs"      "purrr"      "stringr"    "cpp11"      "utf8"      
[21] "stringi"    "pkgconfig"  "withr"      "tools"      "stats"     
[26] "graphics"   "grDevices" 

From a user's point of view, getting simple statistics like OR, RR, or AR from a 2x2 table is not as easy as it could be. Right now, the user needs to know about statistical methods, like using fisher.test for OR. Other contingency table stats are not directly supported, even though they can be calculated easily from a base table().

Would you be open to a lightweight utility or summary function in cardx for these basic 2x2 table stats? Or do you feel this is still outside the package's intended responsibilities? I am happy to follow whatever fits best with the project’s vision.

Thank you again for your insights:D

[ddsjoberg]

If we're including the confidence intervals for these estimates, then I think citools will be a great match (and I agree, it is so much hard to get these estimates than it should be!)

FYI the project will likely be renamed to cicada in the near future

[bzkrouse]

@Polkas please see cicalc here - https://github.com/GSK-Biostatistics/cicalc (prior name citools). Feel free to propose functionality there!