-
Notifications
You must be signed in to change notification settings - Fork 25
Open
Labels
Description
When creating FFTs, FFTrees determines cue thresholds and selects and combines cues based on the goal.threshold, goal.chase and goal parameters.
Although this worked in an earlier version and dprime is still noted as a goal_valid when creating FFTs, actually using it (as any of the goal parameters) currently fails.
As some colleagues favor it as the measure to maximize, a first question is:
- Should we bring back
dprimeas a legitimate measure to maximize?
Reflecting on this, I see two related questions:
- As I'm not sure whether the
$z$ -transformation required to computedprimemakes sense for typical datasets, we could simply maximize the difference of hit rate minus false alarm rate (aka. "column power" or
$\Delta P_c = \text{sens} - (1 - \text{spec})$ , see Table 3 of 10.3389/fpsyg.2020.567817 for details).
This immediately raises the question:
- Is there an analog measure in the orthogonal "prediction direction" that maximizes
the difference between$\text{ppv}$ and$1 - \text{npv}$ ? The answer is yes, of course:
$\Delta P_r = \text{ppv} - (1 - \text{npv})$ ?
Hence, I wonder:
- Has anyone systematically compared the results of those measures with each other?