Hi, this is my attempt at blogging. I will use this space to post ideas and my random thoughts about science, statistics, politics, and share notes about what I am learning (see blog and misc). All views my own, except those nicked from cleverer people.
I am a Lecturer at the Department of Psychology of Royal Holloway, University of London. I use psychophysics, eye-tracking, and computational modelling to investigate visual perception, broadly defined as the ability to assimilate information contained in visible light. Beyond vision, I am interested in how brains compute and use information about uncertainty when making decisions, and in how these abilities changes during development.
Find my publications on Scholar or RG. I started sharing all my code and data on GitHub and OSF.
Some resources (short notes, slides, code) on various topics, mainly statistics.
Multilevel modelling: frequentist and Bayesian approaches. These are the slides for a more advanced course on multilevel modeling. I start by introducing multilevel models from a frequentist perspective, then introduce the Bayesian approach. (The slides are in html format; use arrowkeys to advance) The analyses are performed using Stan and its R interface. Download also the R code that documents the analyses step-by-step (code); the two dataset used in the examples (bisection, mixture); the Stan code of the models discussed (sleepstudy, glmm, mixture). You can also check the Github repository with the Rmarkdown script used to generate the presentation here.
Bayesian model selection at the group level. This note illustrated the principles of Bayesian model selection at the group level, that is treating models as random effects and allowing for individual differences in the model that best describe the behavior of a given participant (see Penny, Daunizeau, Moran, and Friston. Bayesian model selection for group studies. NeuroImage, 46(4):1004–1017, 2009). I implemented (mainly: translated from Matlab, SPM 12) the iterative algorithm and computations to compute exceedance probabilities, and put them in an R package available at this Github repository.
Adaptive maximum likelihood estimation of psychometric slope. Short note illustrating how to set up an adaptive maximum likelihood procedure optimized to estimate the psychometric slope. Contains a closed-form expression of the expected variance of the slope that takes lapse rates into account (thus can be used to compute the so-called sweet points). A Matlab implementation can be found (in additiona to a simple implementation of the Quest+ procedure) in this Github repo.
Linear and generalized linear mixed-effects models in R
. These are the slides for an introductory class on mixed-effects models in R
. Only assume some notions on linear models and some basic experience with R
. Download also: R script; additional dataset, for the last example. Here is also an updated version of part 1 (linear mixed-models only), which I recently presented at the UCL Institute of Ophthalmology.
Eye movements in cognitive sciences (book chapter; in italian). This is an introductory book chapter about the study of eye movements in cognitive sciences. Contains some informations about how they are classified, measured and analysed, and how they can be used to test hypotheses about the mind. The chapter is contained in a volume published by Il Mulino editore, (Title: Il cervello al lavoro. Nuove prospettive in neuropsicologia; ISBN 978-88-15-27211-9), which should be used as textbook at the School of Psychology, in the University of Padova.
Note on Signal Detection Theory and Generalized Linear Models. Short note about the equivalence between equal-variance signal detection theory models and GLMs.