Fits a psychometric function to an individual subject's trial-wise responses using negative log likelihood (nLL)
Right now the function is forced to fit a Weibull, will add more options via a new input argument in the future
- objective function: Weibull
- cost function: nLL
inputs:
- file
{.csv}: file containing trial-wise subject responses (should follow tidy-data format and has to be named'datafile.csv'should have the following column labels{'xvals','trialsIdx','conditions','accuracy'}in no specific order where:'trialsIdx'is the trial indexes eg. 1...nTrials'conditions'is a column with condition labels could be numerial or string or both'accuracy'is the observer's correct (1) or incorrect (0) responses across trials
| trialsIdx | xvals | conditions | accuracy |
|---|---|---|---|
| 1 | 0.05 | valid | 0 |
| 2 | 0.2 | neutral | 1 |
| 3 | 0.85 | invalid | 1 |
-
chance
{float}: what is chance performance in your task? example: if a detection task then there are 2 response alternatives (yes/no) so enter 0.5, if 4 response alternatives enter 0.25 -
plot
{bool}: if True plots the fits / False = no plots
output:
output.npy file in the our_dir containing:
- % correct per condition
- best fit functions per condition
- best fit parameters
CREATED BY: Antonio Fernandez (af) [Oct. 20, 2022]
last edited: Oct. 28, 2020 af