Template talk:DiagnosticTesting Diagram

Latest comment: 5 years ago by Jonmcauliffe in topic Power?

Add accuracy?

edit
Added section title. —Nils von Barth (nbarth) (talk) 05:01, 2 July 2014 (UTC)Reply

Could somebody add Accuracy the bottom right as in this table?


Condition as determined by Gold standard
True False
Test
outcome
Positive True positive False positive Positive predictive value or Precision
Negative False negative True negative Negative predictive value

Sensitivity or recall

Specificity (or its complement, Fall-Out)
Accuracy

What does this show?

edit
 

I don't understand what this diagram shows. What do the different colors stand for? Why are the things in the two right columns grouped together? I understand each of the terms, but I don't think the diagram helps to organize the concepts, at least not as currently designed. Sancho 21:08, 30 March 2016 (UTC)Reply

Hi Sanchom, Their relationships are rather complicated and the formulas are given in the table. The colours attempt to show combinations: e.g. if Z = Y / X, Z's colour is roughly a mixture of Y's and X's. Additionally, true positive and true negative are biased towards green, and false positive and false negative are biased towards red.
I've added this map of relationships, but if it's still confusing, I'd appreciate suggestions to improve it. Thanks, cmɢʟeeτaʟκ 23:56, 28 June 2017 (UTC)Reply
Hi Cmglee, This chart is incredible. It's been incredibly useful to me as I sort through the different measures of accuracy. Would you be open to a couple of ideas to make it more approachable? (right now it's propping up a bunch of highly technical and esoteric pages in which it is the only shining light).
Also, I'm new to Wikipedia, so this may not be the right way or place to communicate. I'm open to moving Cnrmck (talk) 03:35, 15 March 2018 (UTC)Reply
Hi Cnrmck, Thanks for your thoughts. Yes, I'd certainly like to hear your ideas on improving it. I know it's grown messy and much appreciate suggestions to make it clearer while retaining the information.
This is also a good place to keep the conversation going. May I just request for
Hi Cmglee,
to be added to each message to me, as you've done; otherwise I may miss your messages? Thanks, cmɢʟeeτaʟκ 02:09, 18 March 2018 (UTC)Reply

Power?

edit

Hi, I don't see why "Power" appears in the top-left cell of the 2x2 table. In this context, power should correspond to something like (# true positive) / (# condition positive), i.e., "TPR" or "sensitivity" or "recall". It definitely isn't just the number of true positives. In any case, the "power" terminology is common in the context of a formal statistical hypothesis test but rarely used in diagnostic settings (where TPR/sensitivity/recall are used instead).

This table is (a) already jam-packed and (b) an incredibly useful resource, so I hope there is consensus around the small improvement of deleting the word "Power".

Jonmcauliffe (talk) 01:55, 16 January 2019 (UTC)Reply