From our results, adaptive staircases that use fixed step sizes overestimate slope to the same extent as conventional MOCS, whereas adaptive methods that progressively reduce step size to end up placing most trials virtually at a single stimulus level (e.g., stochastic approximation, QUEST, ML-PEST, YAAP, etc) overestimate slope to a much larger extent.
When the goal is estimating all the parameters of [PSI], our results advise against the use of QUEST and similar methods (stochastic approximation, ML-PEST, YAAP, etc) that progressively reduce step size to end up placing most trials within a narrow range of stimulus levels.