OS (e.g. Win10): Win10 PsychoPy version (e.g. 1.84.x): v2021.2.3 Standard Standalone? (y/n) If not then what?: y What are you trying to achieve?:
I’m building an adaptive (QUEST) staircase to assess contrast level for threshold perception (50% hit rate) of a line drawing.
What did you try to make it work?:
I am attaching screenshots of my task here:
What specifically went wrong when you tried that?:
No matter what I try, I get sawtooth results when I plot the intensity changes as a function of trial number.
Maybe I should specify step size? If so, how? Why do intensity changes jump so much in my staircase?
I am less familiar with viewing plotted QUEST intensities - so perhaps I am missing how you are expecting the output to look here (to me the jumping is fine as long as its converging towards a threshold and the stepsizes appear to be adaptively changing - which does lok ok to me there, unless it is the case that you know the upward jumps are not following incorrect responses - in which case that is weird). Could you share your psychopy file and conditions spreadsheet if possible?
Thanks for your reply!
The y-axis should actually be labeled “contrast” since that’s the variable name that I am calibrating using this staircase method.
I decreased my step size to 0.01, since my labmate suggested that a step size of 1 is too big if my contrast value range is between 0 and 1. Still, I get a “sawtooth”. One variable I am unsure about is “range”, since there’s a lot of online discussions about its usefulness in the psychopy quest staircase.
I’ll definitely look at questplus and see if there are other parameters I can play with!
After troubleshooting with my labmate, the issue was with the Weibull function parameters. Partially adapted from the PsychoPy staircase demo:
beta = 3.5
delta = 0.01 gamma = 0 (instead of 0.5; this got rid of the jagged fluctuations of my contrast changes with every step).
grain = 0.01
It turns out that gamma defines the false alarm rate (or, the chance of ‘success’ at an infinitely low stimulus strength). If I set gamma = 0, the jagged fluctuation disappears, because I have a false alarm rate of 0.
So, is this much fluctuation in the last 10 trials of the staircase inherently bad? This looks like I need to do some more reading and understanding of the Weibull function. I’m not sure which would be optimal for my task, yet. If this is helpful to anyone, gamma = 0 parameter is based on advice from this paper: https://link.springer.com/content/pdf/10.3758/BF03201438.pdf (Harvey 1986).
I’ve been trying to make a similar experiment work for me but am unsuccessful. If you don’t mind, could you send me your psyexp file? I would really appreciate it!