PSICOSTAT 3.2

30thApril 2021

Antonio Maffei: An "insider" perspective on (some) methodological and statistical pitfalls in the analysis of electrophysiological data for cognitive neuroscience

Want to know more? Contact Antonio at antonio.maffei@unipd.it

Gianmarco Altoè: A quick tour from linear regression to generalized additive mixed-effects models: an eye to pupillometry data

Want to know more? Contact Gianmarco at gianmarco.altoe@unipd.it

08thApril 2021

Ottavia Epifania: Dalla parte degli item. Georg Rasch e i modelli dell'Item Response Theory

In questa presentazione, verranno affrontati i principali modelli dell'Item Response Theory (IRT) per dati sia dicotomici sia politomici. Si partirà dalle intuizioni di Georg Rasch che hanno portato alla formulazione dell'omonimo modello, secondo cui la probabilità di risposta corretta dipende esclusivamente dalle caratteristiche del soggetto, espresse dalla sua abilità, e da quelle dell'item, espresse dalla sua difficoltà. Verranno poi presentati i modelli IRT che prevedono l'intervento di altre caratteristiche dell'item oltre alla difficoltà nell'influenzare il rapporto tra il tratto latente e la probabilità di risposta corretta. Infine, verranno molte sinteticamente presentati i modelli dell'IRT per i dati politomici. Il codice usato per la presentazione, per la simulazione dei dati e per la creazione della shiny app è disponibile sulla mia pagina GitHub

Please, click here for the PDF of the presentation

Want to know more? Contact Ottavia at otta.epifania@gmail.com

Alberto Arletti: Stima dell'interesse per topic editoriali ispirata all'Item Response Theory

Please, click here for the PDF of the presentation

Want to know more? Contact Alberto at alberto.arletti@studenti.unipd.it

18th December 2020

Prof. Massimiliano Pastore: I dati imputati: colpevoli o innocenti?

In statistica, l'imputazione consiste nell'operazione di sostituire i dati mancanti con dei nuovi valori (non direttamente osservati). Esistono molte tecniche per effettuare questa operazione, disponibili già dalla fine degli anni '80. In questa presentazione vediamo un esempio di imputazione di dati nel contesto di una ricerca longitudinale su bambini tra i 3 e i 12 anni per valutare quali possano essere gli esiti di tale operazione; in particolare cerchiamo di capire se l'imputazione possa essere utile nell'ambito di dati psicologici.

Please, click here for the codes of the presentation.

Want to know more? Contact Massimiliano at massimiliano.pastore@unipd.it

Enrico Toffalini: Design analysis and mixed-effects models. Suggestions on testing treatment efficacy with small effects and small samples

Assessing the efficacy of dyslexia treatments is challenging because the possibility of improvement (i.e., the effect size) is limited, and collecting large sample sizes is difficult. Nearly all published studies are underpowered under any plausible assumption, yet most of them claim treatment effectiveness, generally based on a liberal use of uncorrected multiple testing. This leads to a high risk of false positives and near certain overestimation of the effect sizes that are found statistically significant. To conduct better studies, a strategic use of repeated measurements with mixed-effects modelling is recommended. Increasing the number of observations per participant, even with a limited sample size, allows to crucially enhance power, precision, and reduce the risk of overestimation. Importantly, it also opens the doors to the investigation of the individual differences by considering random effects. Conducting ad hoc, a priori design analyses, fully personalized via simulation to reflect one's assumptions, is strongly recommended.

Please, click here for the presentation slides.

Want to know more? Contact Enrico at enrico.toffalini@yahoo.it

10th November 2020

Filippo Gambarota: A space travel into the tidyverse

Tidyverse is a collection of packages with the aim of improving the readability, reproducibility, and clarity of the data analysis workflow. Each step of data analysis, from importing to data visualization and communication, can be easily performed using a common grammar and style of programming. After a brief introduction about the general principles of the tidy approach, the main packages and functions are presented with some practical examples. Are you ready to switch to a tidy workflow?

Please, click here for the codes of the presentation.

Want to know more? Contact Filippo at filippo.gambarota@gmail.com

Simone D'Ambrogio: The attentional drift diffusion model of multiattribute value based decisions

Want to know more? Contact Simone at simonedambrogio@live.it

27th November 2020

Giulia Calignano: Keep it maximal, be parsimonious: uncertainty dominates the selection of random effects structure in linear mixed-effects models

Non-independent data (as in all cases of a repeated-measures designs) require linear mixed eects modeling (LMEMs). This is a recommended choice, at least for ve very good reasons: (1) violation of the independence assumption lead to incorrect inferential statistics using ANOVA, (2) a single LMEMs can replace two separate ANOVAs, (3) LMEMs better preserve statistical power by dealing with missing data and unbalanced repeats, (4) LMEMs are suited to analyze change over time (or space) higher-order clustered, (5) it is a small step from LMEMs to generalized linear-mixed eects models (GLMEMs), that can also specify the family distribution of residuals to get a more plausible estimate of the models’ parameters. The main characteristic of LMEMs is the random structure. That is, the possibility to specify variance components for each within-subject (S) and within-item eect (I) and interaction term (SX i.e., random slopes). However, the debate about which trade-o between a maximal vs a parsimonious approach ferries to a best practice for a random eects structure remains open. Indeed, it is important to prevent overtting modeling, to avoid inating Type I error and loosing statistical power. This presentation briey review the current state of the art (limited to the perspective of a researcher in cognitive and language sciences), by presenting the coecient estimate comparison between a maximal vs parsimonious modelling of eye-tracking data, by an original repeated measures design.

References
Barr et al., 2013, J. Mem. Lang., https://doi.org/10.1016/j.jml.2012.11.001
Bates et al., 2015, arXiv:1506.04967v2
Brauer Curtin, 2018, Psychol. Method, 10.1037/met0000159
Matuschek et al., 2017, J. Mem. Lang., https://doi.org/10.1016/j.jml.2017.01.001

Please, click here for the presentation.

Want to know more? Contact Giulia at calignano.giulia@gmail.com

Irene Valori: Kinematic measures of inhibition in children with ADHD

The development of motor skills is strictly connected to the optimisation of cognitive abilities, such as the effective inhibition of incorrect or inappropriate responses. Difficulties to inhibit motor behaviours are common to neurodevelopmental disorders such as ADHD (Attention Deficit Hyperactivity Disorder). However, the motor and cognitive processes beneath the profound interindividual differences that characterize this population are still unclear. For instance, motor planning and control might play distinctive roles in the way children execute or inhibit the prepotent response and execute an alternative option. To explore these mechanisms, we applied an adapted version of the Go/No-Go task that required a reaching movement in either a prepotent or alternative condition. On a cohort of 13 children with ADHD (6-13 years old), a low-cost 3-axis wrist worn accelerometer was employed to analyse kinematic measures that discriminate between the planning and the control components of the two movements. We discussed the research design and variables of interest, as well as open questions about the statistical approach that would better address the research questions.

Please, click here for the presentation.

Want to know more? Contact Irene at irene.valori.1@phd.unipd.it