Dear colleagues, I've written a short note relating to a simulation I ran to clarify
my muddy thinking about the effects of bias (towards match or mismatch) in face matching
experiments and the way that principal components analysis separates match and mismatch
items into different components. I can't see it making a published paper but I figure
others may find it useful, so I've put it up on psyarxiv:
https://psyarxiv.com/f2a9j
Bottom line: when participants vary in bias and ability independently, PCA tends to
separate match and mismatch trials, especially after varimax rotation.
The (not very elegant) matlab simulation code is on OSF, linked from the paper.
Comments welcome, to me rather than the whole list.
Peter
Peter Hancock (he/him)
Professor
Psychology, School of Natural Sciences
University of Stirling
FK9 4LA, UK
phone 01786 467675
http://rms.stir.ac.uk/converis-stirling/person/11587
@pjbhancock
Latest paper:
Simulated automated facial recognition systems as decision-aids in forensic face matching
tasks.<https://psycnet.apa.org/record/2023-24366-001?doi=1>
https://psycnet.apa.org/doiLanding?doi=10.1037%2Fxge0001310
My messages may arrive outside of the working day but this does not imply any expectation
that you should reply outside of your normal working hours. If you wish to respond, please
do so when convenient.
________________________________
Scotland's University for Sporting Excellence
The University of Stirling is a charity registered in Scotland, number SC 011159