Uncertain Interval

yurismol · Post by **yurismol** » Tue Oct 21, 2025 4:19 am

Dear Jamoviers!
At the insistent request of @rabouqal, I completed the creation of the Uncertain Interval "Three-way threshold determination of medical tests" procedure in jYS module (R package from Hans Landsheer https://github.com/HansLandsheer/UncertainInterval).
While I was figuring it out, I realized that this is a tool for understanding when making dichotomous decisions.
There were problems with Weblate on Sunday, so this week it only works in English.
The work is not yet finished, so your questions and comments are extremely important.

yurismol · Post by **yurismol** » Tue Oct 21, 2025 4:27 am

Some images

Tak101 · Post by **Tak101** » Wed Apr 22, 2026 7:11 pm

yurismol wrote: ↑Tue Oct 21, 2025 4:19 am Dear Jamoviers!
At the insistent request of @rabouqal, I completed the creation of the Uncertain Interval "Three-way threshold determination of medical tests" procedure in jYS module (R package from Hans Landsheer https://github.com/HansLandsheer/UncertainInterval).
While I was figuring it out, I realized that this is a tool for understanding when making dichotomous decisions.
There were problems with Weblate on Sunday, so this week it only works in English.
The work is not yet finished, so your questions and comments are extremely important.

I found this message, and I am glad there is some interest in my work. Although I discontinued working on this, I still think it is a good idea to use trichotomization for making (medical) decisions based on a test. The uncertain interval shows you the area were a decision would be most uncertain and when a dichotomous decision for or against a disease would be most unwise. Postponing a decision, awaiting further developments, using a better test etc. would be better. The two areas outside the uncertain area will show improved validity (sensitivity/specificity) of the decisions. This method is however NOT a technique to improve on tests that have insufficient sensitivity/specificity. If there are any questions concerning this method, please let me know. Hans Landsheer.

sbalci · Post by **sbalci** » Fri Apr 24, 2026 7:34 pm

can this be used for cytology-biopsy correlation? Usually there are 3 groups of results in cytology: benign, atypical, malignant. and the final biopsy result (gold standart) is benign or malignant. We usually want to assess sensitivity or predictive value of atypical diagnosis, and it becomes hard to understand the result of 3x2 structure.

yurismol · Post by **yurismol** » Sun Apr 26, 2026 1:05 am

This can be used to transform any binary decision into a three-level one: a positive decision, a negative decision, and a third state - uncertainty, that encourages continued exploration before making a final decision.

yurismol · Post by **yurismol** » Sun Apr 26, 2026 1:13 am

Tak101 wrote: ↑Wed Apr 22, 2026 7:11 pm I found this message, and I am glad there is some interest in my work. Although I discontinued working on this, I still think it is a good idea to use trichotomization for making (medical) decisions based on a test. The uncertain interval shows you the area were a decision would be most uncertain and when a dichotomous decision for or against a disease would be most unwise. Postponing a decision, awaiting further developments, using a better test etc. would be better. The two areas outside the uncertain area will show improved validity (sensitivity/specificity) of the decisions. This method is however NOT a technique to improve on tests that have insufficient sensitivity/specificity. If there are any questions concerning this method, please let me know. Hans Landsheer.

I am very pleased with the author's opinion.

jamovi