Talk:Confusion matrix/Operation Transposition

@Danilosilva128: noted on Talk:Confusion_matrix#Confusion_Matrix_is_Transposed_compared_to_standard_practice that confusion matrices in statistics books more commonly have actual (ground truth) values as rows and predicted values as columns, as opposed to the current convention on Wikipedia, and proposes changing Wikipedia articles accordingly.

To avoid inconsistency during the process, @Cmglee: suggests identifying all articles needing changes, so that most (if not all) can be done in a short time, such as during an edit-a-thon.

Please append an asterisk to each entry below if it contains an image requiring modification:

Done Confusion matrix
Done Template:DiagnosticTesting_Diagram
Done Template:SensSpecPPVNPV
- Done transcludes Template:DiagnosticTesting_Example
Done Evaluation_of_binary_classifiers#Contingency_table
Done Matthews_correlation_coefficient#Example
Done Receiver_operating_characteristic#ROC_space
Done Diagnostic_odds_ratio#Example
Done Total_operating_characteristic#Basic_concept
Done Accuracy paradox
Done Prosecutor's_fallacy
- Done File:Prosecutor_fallacy.svg
  File:Prosecutor_fallacy.svg
Done Bayes'_theorem#Examples

User:Danilosilva128 additionally found these on 2022-02-10:

Statistical_hypothesis_testing
Type_I_and_type_II_errors
Durbin–Wu–Hausman_test
Base_rate_fallacy
Multiple_comparisons_problem
Family-wise_error_rate
Prosecutor's_fallacy
False_discovery_rate
False_coverage_rate
Detection_theory (inconsistent ordering of positive/negative)

and these on 2022-02-11:

(all these new additions were based on searches for "statistics/statistical + test") — Preceding unsigned comment added by Danilosilva128 (talk • contribs) 15:29, 11 February 2022 (UTC)[reply]

The following articles need not be changed, as they already use the proposed convention (inconsistently with the ones above):

Many related articles need not be changed since they either use a template listed above or do not give a specific example of a confusion matrix.

Note: I covered all articles in the search for "confusion matrix", "table of confusion" and "contingency table". In the latter search, the articles either refer to a generic contingency table or already mention the keyword "confusion matrix". So I believe these are all the articles that need to be changed.

@Cmglee:, since these are few articles, I think this can be done by the two of us. For instance, we decide the changes previously, and then a single person edits all the articles and clicks "Publish changes" at approximately the same time on all of them. What do you think? (I have no experience with edit-a-thon) Danilosilva128 (talk) 16:17, 8 May 2021 (UTC)[reply]

@Danilosilva128: Thanks a lot for identifying the articles. That looks quite manageable and I'll be glad to do it tomorrow. The big one are the two templates Template:DiagnosticTesting_Diagram and Template:SensSpecPPVNPV. Since I worked on them before, shall I take them?

Curiously enough, there's a London meetup tomorrow. The last one I attended (virtually), we edited several articles together, pooling in resources and sharing tips. I've proposed it to see if anyone is interested. Cheers, cmɢʟee⎆τaʟκ 16:49, 8 May 2021 (UTC)[reply]

@Cmglee: Sure, please take the two templates (and any other you wish). If no one is interested in the remaining ones, I can take them. Just let me know.

@Danilosilva128: Thanks, I've created template:diagnostic testing diagram and template:diagnostic testing example so that their respective transclusions be switched over during the London meetup at 12:00 UTC (13:00 BST, 14:00 CET). I'll put a

Done against those I've finished as above. Cheers, cmɢʟee⎆τaʟκ 10:17, 9 May 2021 (UTC)[reply]

P.S. I'll update the transclusions now. cmɢʟee⎆τaʟκ 12:36, 9 May 2021 (UTC)[reply]

@Danilosilva128: I've done all the pages you've identified. When you get online, can you please check if I've missed anything? Thanks, cmɢʟee⎆τaʟκ 14:04, 9 May 2021 (UTC)[reply]

@Cmglee: Thanks for being so quick! I checked all the articles and fixed the few things left that I noticed. It is all looking pretty good now. Thanks so much for helping with this! Danilosilva128 (talk) 14:44, 9 May 2021 (UTC)[reply]

My pleasure. Thanks for initiating the idea. May I get your opinion on one thing? Several of the articles using Template:diagnostic testing diagram introduce it with "Let us define an experiment from P positive instances and N negative instances for some condition. The four outcomes can be formulated in a 2×2 contingency table or confusion matrix, as follows:", though the diagram doesn't actually use "P" and "N" and the table is more than 2×2. Can you suggest a better way to introduce it? Cheers, cmɢʟee⎆τaʟκ 20:15, 9 May 2021 (UTC)[reply]