Additions:
By default, a psychometric function ranging from 0% to 100% is used, as
is suitable for discrimination experiments with a standard in the middle
of the possible stimulus parameter range. For other paradigms, such as
n-AFC detection tasks, one can set the guessrate input during staircase
init to 1/num_alternatives, e.g. .5 when doing a 2IFC detection task.
This guess rate is thus not the rate at which participants guess instead
of do your task (thats the lapse rate), it the minimum rate of correct
responses as determined by your design. NB: below discussion is based on
the default psychometric function with the full range, but all points are
equally valid for a scaled psychometric function.
is likely, because it's near the correct 25% point), then for the next
There is an option to set the first probe value to be tested, which, for
the first trial only, will overrule both of the above probe choice
strategies. This can be useful if you want to be sure that the first
trial is an easy one so the participant knows what to expect.
Another measure for robustness is to choose a small lapse rate. If lapse
whole range of mean-slope combinations becomes impossible. If lapse rate
staircase can rebound. Therefore a lapse rate of 5% or even more
depending on task difficulty is always recommended. NB: in the default
discrimination setup of the staircase (guessrate is not specified or set
to 0), half of the lapse rate is taken off the bottom of the psychometric
function and half is taken off the top. So if the lapse rate is 0.05, the
psychometric function will range from 0.025 to 0.975. In the setup for a
n-AFC detection experiment when the psychometric function has a lower
bound of 1/num_alternatives, the lapse rate is taken off the top. So when
the guess_rate is set to .5 (2AFC) and the pase rate is set to .05, the
psychometric function will range from 0.05 to 0.95.
Note that the staircase does not support a 0 lapse rate in the first
take the log of a 0 probability. Any lapse rate lower than 1e-10 will be
lapse rate is assumed, it is common to either fit the psychometric
function without a lapse rate, or otherwise with the lapse rate as a free
Saunders JA & Backus BT (2006). Perception of surface slant from
oriented textures. Journal of Vision 6(9), article 3
of the different staircase functions.
Deletions:
is likely, because it's near the correct 25% points), then for the next
Another measure for robustness is to choose a small guess rate. If guess
whole range of mean-slope combinations becomes impossible. If guess rate
staircase can rebound. Therefore a guess rate of 5% or even more
depending on task difficulty is always recommended.
Note that the staircase does not support a 0 guess rate in the first
take the log of a 0 probability. Any guess rate lower than 1e-10 will be
guess rate is assumed, it is common to either fit the psychometric
function without a guess rate, or otherwise with the guess rate as a free
Saunders JA & Backus BT (2006). Perception of surface
slant from oriented textures. Journal of Vision 6(9), article 3
of the different staircase functions