Meetings' calendar

Seventh Meeting

25th June 2019

Giovanni Frighetto: Visuomotor control of perturbed goal in freely walking Drosophila

The mechanism of action selection is a widely shared fundamental process required by animals to interact with the environment and adapt to it. A key step in this process is the filtering of the “distracting” sensory inputs which may disturb action selection. Because it has been suggested that, in principle, action selection may also be processed by shared neural circuits in vertebrates and invertebrates I wondered whether invertebrates show the ability to filter out “distracting” stimuli during a goal-directed action, as seen in vertebrates. In this meeting I presented how action selection might be implemented by fruit flies, investigating their reaction to the abrupt appearance of a visual distractor during an ongoing locomotor action directed to a visual target, in terms of adopted paths. A simulation of the trajectories based on the retinal slip motion compensation has only partially explained the flies’ behaviour leaving open the issue related to the action selection mechanisms.

Want to know more? Contact Giovanni at

Prof. Massimiliano Pastore: Introduction to STAN

Prof. Gianmarco Altoè: Using truncated normals to formalize plausible effect size

Sixth Meeting

11th June 2019

Prof. Franca Agnoli: Questionable Research Practices

Discussion on Questionable Research Practices and presentation of the paper "Questionable research practices among italian research psychologists"

Click here to see the paper.

Giulia Bertoldo: Registered Report - Improving decision-making in the criminal justice system: The effect of teaching aids on the evaluation of eyewitness evidence

Eyewitness testimony is one of the most compelling types of evidence used in trials but these testimonies are not always accurate and have led to wrongful convictions in the past. Judges, jurors and other triers of fact rely heavily upon eyewitnesses. However, lack of awareness of evidence-based facts as well as sources of error in eyewitness testimony can impair decision-making. Effective communication with triers of fact is therefore of utmost importance in achieving reliable and fair trials. This project will test the effects of communicating scientific guidelines to triers of fact in legal decision-making settings. We aim to build upon existing communication procedures to try to improve fairness in the criminal justice system.

Want to know more? Contact Giulia at

Fifth meeting

Fifth Meeting

30th May 2019

Tania Moretta: Understanding gender differences in recovery from gambling disorder: a Bayesian adaptive sampling for variable selection and model averaging

Understanding gender-related differences is important in recovery processes. Prior studies have investigated gender-related differences in factors associated with gambling disorder (GD), but none to date have considered both positive and negative resources related to recovery. Using a recovery-capital (RC) framework that considers multiple resources available during recovery, this study explored gender-related similarities and differences in associations between positive (RC, spirituality) and negative (stressful life-events, depression, anxiety) resources and GD symptom improvement on one-hundred-and-forty individuals with lifetime GD (101 men). Multicollinearity was monitored by examining both the tolerance and the variance inflation factor (VIF). Residual plots were employed to evaluate the normality and homogeneity of variance. A Bayesian approach was employed to explore whether RC, spirituality, stressful life-events, depression, and anxiety and their two-way interactions provided adequate descriptions of the distributions generating the observed GD symptom improvement in women and men, separately. Specifically, a Bayesian adaptive sampling for variable selection and model averaging was employed. All possible combinations between predictors were estimated by a Markov chain Monte Carlo (MCMC) sampling method using the Zellner-Siow Cauchy prior on the coefficients and a uniform prior distribution over the models. As an extension of Bayesian inference, this approach provides a coherent and systematic mechanism that accounts for model uncertainty during the process of variable selection by considering both parameter uncertainty through the prior distribution and model uncertainty, and obtains posterior distributions for the model parameters and the model themselves using Bayes’ theorem, hence allowing for direct model selection, combined estimation and prediction. Finally, to test regression slope homogeneity between genders we computed the Bayes factor (BF) by 100,000 MCMC simulations for all restrictions of the full model (i.e. including GD symptom improvement as the dependent variable and the statistical predictors that were selected from previous analyses and were in common across gender groups, and their interaction with gender as independent variables) against the null hypothesis that all effects were 0. Findings showed that RC is an important positive resource for men and women recovering from GD and should be considered in treating both groups. Understanding specific RC factors across gender groups and stressors in men may aid in developing improved interventions for GD.

Want to know more? Contact Tania at

Letizia Dalla Longa: Comparison between statistical models to study infants’ looking behaviour

In developmental research, looking behaviour provides an important measure to investigate infants’ attention and cognitive processing. In particular, visual preference paradigm consists in presenting two side-by-side stimuli and measuring how long infants look to each of them. This paradigm can provide information about infants’ ability to discriminate between two stimuli and to spontaneously prefer one to another. Which statistical model should we use when analysing looking times? How can we take into account individual variability?

Want to know more? Contact Letizia at

Giulia Bertoldo: Design analysis: prospective and retrospective

A design analysis (Gelman & Carlin, 2014) can be interpreted as a broader definition of a power analysis that includes also other two types of errors: Type M error and Type S error. Type M (magnitude) error (or exaggeration ratio) gives information about “how many times” a statistically significant effect size overestimates the plausible true effect size. Type S (sign) error is the probability that a statistically significant result has the opposite sign of the plausible true effect size. A prospective design analysis can be performed at the planning stage of a study in order to choose an adequate sample size and minimize those risks. A retrospective design analysis can instead be performed after a study is conducted, to analyze the extent to which these risks are present in a study.

Want to know more? Contact Giulia at

Fourth meeting

Fourth Meeting

16th May 2019

Laura Giuntoli: Network modeling and psychological constructs dimensionality

Network modeling is based on the estimation of the weight matrix that depicts the relationships between observed variables (i.e., items of a questionnaire) as partial correlation coefficients. The strength is a measure of centrality that represents the influence of each observed variable on the others and it is obtained by the row sums in the weight matrix. To describe the dimensionality of a psychological construct, a decomposition of the strength index is presented.

Want to know more? Contact Laura at

Simone D'Ambrogio: Estimation of the attentional drift diffusion model parameters via Random Utility

When people make a decision between a limited set of options, they scan the alternatives with their eyes until eventually they make a choice. A wealth of recent literature illustrates that the attention influence the decision process. A statistical model used to study the role of attention in the two-alternative forced decision process is the attentional Drift Diffusion Model (aDDM). This presentation focuses mainly on approaches that can be used to estimate the aDMD parameters. In particular, the problem of how the random utility models (RUM) can replace the grid-search approach is discussed.

Click here to see the presentation.

Want to know more? Contact Simone at

Anna Giorgia Carollo: Lack of replication in Psychology: from crisis to renaissance

Scientific research, including psychology, is the basis of the actions that are implemented every day to improve the life of people in the world. Science is an approach to acquire and evaluate knowledge through repeated tests. In particular it is an approach that aims to reduce the error by implementing methodological safeguards in order to determine the associated uncertainty (Lilienfeld, 2010). For the past ten years, Psychology has been tackling a replication crisis: as Open Science Collaboration (2015) declares, many psychological findings cannot be replicated, which leads to the growing concern that many published findings are excessively reassuring or even false (almost 50% of published results according to Ioannidis, 2005). Several large-scale problems affect the validity and reproducibility of psychological research. At the origins of the replication crisis some authors have criticized the misuse of the paradigm of statistical significance (Null Hypothesis Significance Testing, NHST) as insufficiently intuitive and engaging researchers to misunderstand the value of p and make statistical errors in published reports (Cumming et al., 2007; Hubbard, 2016; Kline, 2013; Wagenmakers, 2007). In fact, in the psychological literature, most studies do not often report all conditions and measures (Franco, Malhotra, & Simonovits, 2016) nor outcome measures and reasons for terminating data collection (LeBel et al., 2013). Moreover, exploratory analysis is sometimes reported as confirmatory (Wagenmakers et al., 2012). The inconsistencies in statistical reporting are one aspect of the problems that currently afflict psychological science (Bakker & Wicherts, 2011). The high prevalence of statistical errors in psychology papers is worrying and there is evidence that this is not only a problem for psychology. Similar inconsistency rates have been found in the medical sciences (Garcia-Berthou & Alcaraz, 2004). In psychology these errors display a systematic preference for statistically significant results. There are several possible causes for this systematic error prevalence: The File Drawer Problem (Rosenthal, 1979), called also Publication Bias, which arises whenever the probability of a study being published depends on the statistical significance of the results (p < .05), may be the most notable. In fact, as noted by Fanelli (2010) and Ioannidis (2005) the studies published in journals often support a psychological theory (almost 90% of the times) rather than refute it even if the theory in real life is false. This is the reason some authors have made great efforts to understand the research practices that undermine the quality of scientific research in numerous disciplines, going beyond investigating the unethical behavior universally known as FFP (Fabrication, Falsification and Plagiarism) and focusing on research practices defined as problematic and or questionable that fall into an ethical “gray zone” (Butler, Delaney, & Spoelstra, 2017: De Vries, Anderson, & Martinson, 2006). There have been an increasing number of suggestions regarding how to solve these problems. Many authors have proposed changes in statistical inference practices and attribute greater emphasis on methodological rigor and demand better evidence to support strong claims (Mcshane & Böckenholt, 2017; Smaldino & McElreath, 2016; van Aert, Wicherts, & van Assen, 2016). Anyway, whatever’s the solution or the most suitable solutions to overcome the replicability crisis, it is essential that any researcher will be guided by solid and shared ethical principles, such as those proposed by Merton (1942): universality, collegiality, disinterest and organized skepticism.

Want to know more? Contact Anna Giorgia at

text mining

Third Meeting

2nd May 2019

Niccolò Polo: Text mining and gambling TV spots

Gambling tv spots are a relative new field where Psychology can look at in order to understand how gambling suppliers tempt consumers to approach a risky behaviour. Text mining and Natural Language Processing tecniques are useful tools to investigate gambling marketing language. The research focus its attention on three different categories of tv spots: sport betting, casino's type games and bingo/scratch cards. Text mining tools are used to explore and compare this three categories, and then exctract latent informations inside tv spot texts.

Want to know more? Contact Niccolò at

second meeting

Second meeting

11th April 2019

Today discussions spanned from applied statistics in the analysis of fMRI data to cognitive psychology and specific learning disorders! Check out the summaries of our presenters' work!

Angela Andreella: A Statistical approach to the alignment of fMRI data

Multi-subject functional Magnetic Resonance Image (fMRI) studies are critical to test the validity of findings across subjects. However, the anatomical and functional structure varies across subjects, hence the image alignment is a fundamental step. One anatomical alignment is the Talairach Atlas, thus, it doesn't account for functional topography. For that, Haxby et al. (2011) developed a functional approach called Hyperalignment, using sequential Procrustes orthogonal transformations. The inter-subject classification of functional response improved. However, any constraint isn't imposed on the transformation, losing results interpretability. In this contribution, we tackle the functional alignment with a statistical perspective. A probabilistic model for the data generating process is defined. The maximum likelihood estimates of the rotation parameters result to be the Generalized Procrustes Problem (GPA), which improves the Hyperalignment in terms of residuals sum of squares. The statistical framework allows assuming a prior distribution of the orthogonal transformation parameter, as the matrix Fisher Von Mises distribution. It embeds the anatomical information in the estimation of the parameters, i.e. penalizing the combination of spatially distant voxels. The application to several datasets shows that the proposed method improves the classification accuracy of task-related images with respect to the anatomical alignment and the Hyperalignment methods.

Want to know more? Contact Angela at

Enrico Toffalini: "Core deficits" in Specific Learning Disorders: Is it better a dimensional or a taxonomic approach?

Specific Learning Disorders (SLD) are defined by low performance in specific (but not general) areas of cognitive and learning functioning. Although many researchers state that they adopt a dimensional approach to SLD, they generally conduct studies that reflect a categorical (taxonomic) approach. Much research is devoted to identify "core deficits" in basic cognitive/neuropsychological processes, that would uniquely identify individuals with severe difficulties in specific areas. The general statistical question is whether what we observe in a population subset, defined by specific (and somehow arbitrary) rules and cut-offs, can be inferred from the rest of the population, which relates to the broader statistical question of whether a group is part of a population or not. A simulation approach is presented to try to address this question.

Want to know more? Contact Enrico at or at

Fillippo Gambarota & Thomas Quettier: Predominance Ratio in Binocular Rivalry

In binocular rivalry (BR) paradigm two dissimilar images compete and an alternation in perceptual dominance occurs. The two images are constantly processed by the visual cortex. A dominant monocular image has access to the awareness while the other is suppressed. For each eye’s stimulus one may further calculate the predominance as the total proportion of the monocular image that accesses to the awareness. Statistics analysis in BR is made on a Predominace ration (PR). PR is the predominance of one of the two percepts minus the predominance of the other percepts divided by the sum of the predominance of both the percepts. Is PR the best index to use? What about multinomial logistic regression?

Want to know more? Contact Thomas at or Filippo at

first meeting

Psicostat is back!

28th March 2019

Today the psicostat group started PSICOSTAT 3.0, the third edition of a series of meetings where we discuss applied statistics to psychological research. We talked about diffusion decision models and explore the world of Shiny Apps. Our presenters were Simone and Claudio. Below you can find more about their presentations!

Simone D'Ambrogio: Diffusion Models for Simple Decisions

The Diffusion decision model is a statistical model that has been used increasingly in cognitive psychology to study the processes underlying dichotomic decision making. The theoretical assumption underpinning the Diffusion decision model, that simple decisions are based on stochastic processes, came from some neuropsychological evidences. These show how the neural activity of particular brain areas of monkeys which are involved in decision-making tasks, reflects diffusion mechanisms. Of particular interest is the role that the attention plays in this process. Understanding to what extent and how the attention influences the decision mechanisms, would shed more light on the whole decision computational architecture.

Want to know more? Contact Simone at

Claudio Zandonella Callegher: Shiny: Interactive web applications

Description of the main features of Shiny with basic instructions on how to create Shiny apps and some example of applications.

Want to know more? Contact Claudio at