BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:An approach to structural equation modeling in a multiblock framew
ork
DTSTART;VALUE=DATE-TIME:20230630T111000Z
DTEND;VALUE=DATE-TIME:20230630T113000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1035@cern.ch
DESCRIPTION:Speakers: Rosaria Romano (University of Naples Federico II)\nI
n many application fields\, the variables used to measure a phenomenon are
gathered into homogeneous blocks that measure partial aspects of the phen
omenon. For example\, in sensory analysis\, the overall quality of product
s may depend on the taste and odor variables\, etc. In consumer analysis\,
consumer preferences may depend on physical-chemical and sensory variable
s. In some contexts\, a structure of relations between the different block
s may exist that gives rise to a chain of influences. Within each link\, t
he blocks of predictor variables are called input blocks\, while the block
of dependent variables is called the output block. If the input blocks do
not depend on any other block\, then they are defined as exogenous blocks
\, while those that rely on other input blocks in the same relation are ca
lled intermediate blocks. If there is a chain of the relationship between
the blocks\, we are then dealing with what is often called a mediation mod
el and must interpret both indirect and direct effects among blocks.\nWith
in the scope of multiblock data analysis with a directional path among the
blocks\, we will present a new approach named SO‐PLS path modelling (SO
‐PLS‐PM). \nThe approach splits the estimation into separate sequentia
l orthogonalized PLS regressions (SO-PLS) for each output block. The new m
ethod is flexible and graphically oriented and allows for handling multidi
mensional blocks and diagnosing missing paths. New definitions of total\,
direct\, indirect\, and additional effects in terms of explained variances
will be proposed\, along with new methods for graphical representation. \
nIn this research\, some interesting properties of the method will be show
n both on simulated and real data. The actual data concerns consumer\, sen
sory and process modelling data. Results will also be compared to those of
alternative path modelling methods.\n\nKeywords: path analysis\, graphica
l modelling\, multiblock regression\n\n\nReferences\nR. Romano\, O. Tomic\
, K.H. Liland\, A. Smilde\, T. Næs (2019). A comparison of two PLS‐base
d approaches to structural equation modeling. Journal of Chemometrics\, 33
(3)\, e3105.\nT. Næs\, R. Romano\, O. Tomic\, I. Måge\, A. Smilde\, K.H
. Liland. Sequential and orthogonalized PLS (SO‐PLS) regression for path
analysis: Order of blocks and relations between effects. Journal of Chemo
metrics\, 35 (10)\, e3243.\n\nhttps://indico.unina.it/event/67/contributio
ns/1035/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1035/
END:VEVENT
BEGIN:VEVENT
SUMMARY:MCMC or Reservoir computing? A direct sampling approach
DTSTART;VALUE=DATE-TIME:20230630T090000Z
DTEND;VALUE=DATE-TIME:20230630T094000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1002@cern.ch
DESCRIPTION:Speakers: Petros Dellaportas (AUEB and UCL)\nAssume that we wo
uld like to estimate the expected value of a function f with respect to a
density π by using an importance density function q. We prove that if π
and q are close enough under KL divergence\, an independent Metropolis sam
pler estimator that obtains samplers from π with proposal density q\, enr
iched with a variance reduction computational strategy based on control va
riates\, achieves smaller asymptotic variance than the one from importance
sampling. We illustrate our results in challenging option pricing problem
s that require Monte Carlo estimation. Furthermore\, we propose an automat
ic sampling methodology based on adaptive independent Metropolis that can
successfully reduce the asymptotic variance of an importance sampling esti
mator and we demonstrate its applicability in a Bayesian inference problem
s.\n\nhttps://indico.unina.it/event/67/contributions/1002/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1002/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sparse Hierarchical Vector Autoregression for Psychopathological N
etwork Estimation from Intensive Longitudinal Data
DTSTART;VALUE=DATE-TIME:20230630T083000Z
DTEND;VALUE=DATE-TIME:20230630T085000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1013@cern.ch
DESCRIPTION:Speakers: Spyros Balafas (Vita-Salute San Raffaele University)
\nThe use of networks as a tool for studying complex\nsystems gained popul
arity in various scientific disciplines. In the past decade\, the ``networ
k takeover'' reached psychology\, and networks were utilized to abstract c
omplex psychological phenomena. In psychopathology\, a network-based frame
work known as the *network theory of mental illness*\, posits that mental
disorders emerge as systems of causally interacting psychopathological sym
ptoms. According to this framework\, symptoms and other psychological or s
ociological factors are nodes in a psychopathological network\, and the ab
sence of an edge between two nodes corresponds to a conditional independen
ce relationship. In contrast to other types of networks (e.g.\, social net
works) where the structure is observed\, in networks from psychopathology
the dependence structure between the nodes is not known a priori and needs
to be estimated from data. \n\nTypically\, after estimating a psychologic
al network\, summary statistics are used to describe its structural proper
ties both at the global and local levels. In the psychological literature\
, clinical outcomes such as illness severity have been associated with net
work summary statistics such as network density. The aim of this study is
to test i) whether the network density differs across populations of incre
asing illness severity\, and ii) whether local network statistics can be u
sed to identify symptoms that are associated with illness severity. For th
is purpose\, we use intensive longitudinal data from a $90$-day diary stud
y called Mapping Individual Routes of Risk and Resilience (MIRORR). The da
ta consists of $8640$ observations within $N = 96$ individuals\, divided o
ver four subgroups representing different early clinical stages ($n_1 = 25
$\, $n_2 = 27$\, $n_3 = 24$\, $n_4 = 20$). Participants in the lowest risk
group were randomly selected from the general population in the north of
the Netherlands based on their score on the Community Assessment of Psychi
c Experiences (CAPE) test. Inclusion criteria for the study were: aged bet
ween $18$ and $35$ years\, reading and speaking Dutch fluently\, being cap
able of following the research procedures\, provide informed consent. Excl
usion criteria for participating in the study were: psychotic episodes (cu
rrent or in the past) according to the Diagnostic and Statistical Manual o
f Mental Disorders 4 (DSM-4)\, hearing or visual problems\, and pregnancy.
Participants were excluded from the study when they missed more than $22$
measurements in total or missed five or more measurements in a row. Items
in the diary assessment covered a wide range of feelings and (subclinical
) psychotic experiences\, depression\, anxiety\, mania\, obsessive-compuls
ive behavior\, and anger. Participants were required to complete a digital
questionnaire on psychopathological symptoms\, emotions\, functioning\, a
nd stress once a day for $90$ consecutive days. \n\nFor estimating the net
work structure for each group of participants\, we propose a hierarchical
extension of the graphical vector autoregressive (GVAR) model that aims to
model the heterogeneity in intensive longitudinal data. The parameters of
the proposed hierarchical GVAR model are estimated within a two-step proc
edure that combines penalized linear mixed models with graphical LASSO (gL
ASSO). The estimated networks are then used to calculate global and local
network statistics\, which are compared across groups using statistical te
sts. \n\nOur results showed that global network statistics such as network
density and connectivity do not significantly differ as mental illness be
comes more severe. However\, we propose the use of local network character
istics such as centrality indices to identify emotions that correlate sign
ificantly with increasing illness severity.\n\nhttps://indico.unina.it/eve
nt/67/contributions/1013/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1013/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Sparse Pairwise Likelihood Inference for Multivariate Time Series
Models
DTSTART;VALUE=DATE-TIME:20230630T073000Z
DTEND;VALUE=DATE-TIME:20230630T075000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1003@cern.ch
DESCRIPTION:Speakers: Xanthi Pedeli (Athens University of Economics and Bu
siness\, Department of Statistics)\nMultivariate time series data is becom
ing an increasingly common research topic. Unlike univariate time series\,
the temporal dependence of a multivariate series includes both serial dep
endences and interdependences across different marginal series. Consequent
ly\, as the number of component series increases\, multivariate time serie
s models become overparameterized. In addition\, there are many cases wher
e the conditional distribution of the multivariate series given its past m
ight have a complicated form. Given these challenges we develop methodolog
y by replacing the full likelihood function by a pairwise likelihood that
only requires the specification of bivariate marginals instead of the mult
ivariate distribution. Clearly\, the computational task of maximization of
the pairwise likelihood is much simpler than maximization of the full lik
elihood function but still it poses the problem of combining all estimator
s. For this purpose\, we rely on maximization of an approximate weighted l
east squares estimation criterion subject to a shrinkage penalty that allo
ws for model selection. The suggested approach provides a general framewor
k for multidimensional time series since it can be applied to both continu
ous and discrete time series but also to mixed mode time series data.\n\nh
ttps://indico.unina.it/event/67/contributions/1003/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1003/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Semiparametric regression for competing risks data with missing no
t at random cause of failure
DTSTART;VALUE=DATE-TIME:20230629T132000Z
DTEND;VALUE=DATE-TIME:20230629T134000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1024@cern.ch
DESCRIPTION:Speakers: Giorgos Bakoyannis (Athens University of Economics a
nd Business)\nThe cause of failure in cohort studies that involve competin
g risks is frequently incompletely observed. Failure to deal with this iss
ue can lead to substantially biased estimates. To the best of our knowledg
e\, all the methods that have addressed the issue in the context of semipa
rametric competing risks models rely on a missing at random (MAR) assumpti
on. Nevertheless\, the MAR assumption is not realistic in many real-world
settings. In this work we relax the latter assumption by allowing for a cl
ass of missing not at random (MNAR) mechanisms\, which contain the MAR mec
hanism as a special case. Due to the inherent non-identifiability issues u
nder MNAR\, we propose an approach for hypothesis testing that does not re
quire the estimation of the non-estimable parameters. Using modern empiric
al process theory\, we show that the proposed estimators are uniformly con
sistent under the assumed class of MNAR mechanisms. We also show that our
estimators converge weakly to tight zero mean Gaussian processes and propo
se rigorous methodology for the computation of confidence intervals which
achieve a coverage rate of at least 100*(1 - α)%\, asymptotically\, for t
he true unknown parameters of interest. The proposed methodology is applie
d to competing risks data from a large multicenter HIV study in sub-Sahara
n Africa where a substantial portion of causes of failure is missing not a
t random.\n\nhttps://indico.unina.it/event/67/contributions/1024/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1024/
END:VEVENT
BEGIN:VEVENT
SUMMARY:On the predictability of a class of ordinal data models
DTSTART;VALUE=DATE-TIME:20230629T140000Z
DTEND;VALUE=DATE-TIME:20230629T142000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1008@cern.ch
DESCRIPTION:Speakers: Rosaria Simone (University of Naples Federico II)\nT
he contribution aims at discussing some preliminary results on the evaluat
ion of prediction performance for the class of mixture models with uncerta
inty (Piccolo and Simone\, 2019). The ultimate goal of the analysis is the
evaluation of the extent by which the uncertainty specification constitut
es an added value for prediction of ordinal scores. A small simulation stu
dy is presented to assess prediction performance of competing models under
miss-specification. The Ranked Probability Score is chosen as scoring rul
e since it is the most suited to deal with ordinal data\, without the assi
gnment of numerical scores to category. Finally\, a variable selection pr
ocedure based on prediction performance can be outlined on a case study fo
r the prediction of subjective probability to survive. Comparisons with cu
mulative link models are illustrated for the sake of completeness. Prelimi
nary findings discussed in Simone and Piccolo (2022) indicate that uncerta
inty modelling improves prediction performance substantially. Hence\, it i
s important to assess the information quality of the baseline preference m
odel (the Binomial\, for instance). To this aim\, we introduce a new utili
ty measure for preference models when contaminated with alternative uncert
ainty specifications in the sense proper to the framework of Information Q
uality. As a result\, the mixing weight of the chosen feeling component wi
thin the mixture can be explicitly interpreted in terms of model predictiv
e ability.\n\n**Keywords**: CUB models\; Predictability\; Ranked Probabili
ty Score\; Ordinal Data\n\n**References**:\nD. Piccolo\, R. Simone (2019).
The class of CUB models: statistical foundations\, inferential issues and
empirical evidence. STATISTICAL METHOD AND APPLICATIONS\, Volume 28\, pag
es 389-435.\nR. Simone and D. Piccolo (2022). On the predictability of a c
lass of ordinal data models. In A. Balzanella\, M. Bini\, C. Cavicchia\, a
nd R. Verde\, editors\, Book of short papers SIS 2022\, 51st Scientific Me
eting of the Italian Statistical Society\, pages 1053–1058. Pearson.\n\n
https://indico.unina.it/event/67/contributions/1008/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1008/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Maximum likelihood estimation of multivariate regime switching Stu
dent-t copula models
DTSTART;VALUE=DATE-TIME:20230629T105000Z
DTEND;VALUE=DATE-TIME:20230629T111000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1025@cern.ch
DESCRIPTION:Speakers: Federico Cortese (Università degli studi di Milano-
Bicocca)\nWe propose a novel estimation method for multivariate regime swi
tching models based on a Student-t copula function. These models account f
or the interdependencies between multiple variables by considering the cor
relation strength controlled by specific parameters. Moreover\, they addre
ss fat-tailed distributions through the number of degrees of freedom. Thes
e parameters\, in turn\, are governed by a latent Markov process.\n\nWe co
nsider a two-steps procedure carried out through the Expectation-Maximizat
ion algorithm to estimate model parameters by maximum likelihood. The prim
ary computational challenge lies in estimating both the matrix of dependen
ce parameters and determining the number of degrees of freedom for the Stu
dent t-copula. To address this\, we introduce a new approach that leverage
s Lagrange multipliers\, simplifying the estimation process.\n\nThrough a
comprehensive simulation study\, we demonstrate that our estimators posses
s desirable properties in finite samples. Additionally\, the estimation pr
ocedure shows good computational efficiency.\n\nWe apply our model to anal
yze the log-returns of five different cryptocurrencies. The results enable
us to identify distinct bull and bear market periods based on the intensi
ty of correlations observed between the crypto assets. This finding highli
ghts the model's efficacy in capturing and characterizing market dynamics
within the cryptocurrency domain.\n\nKeywords: statistical models for fina
ncial analysis\, cryptocurrencies\, time series\, Expectation-Maximization
algorithm\, latent variable models\n\nhttps://indico.unina.it/event/67/co
ntributions/1025/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1025/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Capturing Correlated Clusters Using Mixtures of Latent Class Model
s
DTSTART;VALUE=DATE-TIME:20230629T113000Z
DTEND;VALUE=DATE-TIME:20230629T115000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1020@cern.ch
DESCRIPTION:Speakers: Gertraud Malsiner-Walli (Vienna University of Econo
mics and Business)\nLatent class models rely on the conditional\n independ
ence assumption\, i.e.\, it is assumed that the categorical\n variables ar
e independent given the cluster memberships. \n Within the Bayesian framew
ork\, we propose a suitable specification of\n priors for the latent class
model to identify the clusters in\n multivariate categorical data where t
he independence assumption is not\n fulfilled. Each cluster distribution
is approximated by a latent\n class model\, leading overall to a mixture o
f latent class models.\n The Bayesian approach allows to identify the clus
ters and fit their\n cluster distributions using a one-step procedure. We
provide suitable estimation and inference methods for the\n mixture of lat
ent class models and illustrate the performance of this\n approach on arti
ficial and real data.\n\nKeywords: Bayesian inference\, model-based cluste
ring\, prior on the number of components\, telescoping sampler.\n\nFop\, M
.\, K. M. Smart\, and T. B. Murphy (2017). Variable selection for latent\n
class analysis with application to low back pain diagnosis. The Annals of\
nApplied Statistics 11 (4)\, 2080-2110.\n\nFruehwirth-Schnatter\, S.\, G.
Malsiner-Walli\, and B. Gruen (2021). Generalized\nmixtures of finite mixt
ures and telescoping sampling. Bayesian\nAnalysis 16 (4)\, 1279–1307.\n\
nMalsiner-Walli\, G.\, S. Fruehwirth-Schnatter\, and B. Gruen (2017). Iden
tifying\nmixtures of mixtures using Bayesian estimation. Journal of Comput
ational\nand Graphical Statistics 26 (2)\, 285–295.\n\nhttps://indico.un
ina.it/event/67/contributions/1020/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1020/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Latent Feeling and Uncertainty of Perception and Expectations of P
rice levels over time: A Change Point Analysis
DTSTART;VALUE=DATE-TIME:20230629T111000Z
DTEND;VALUE=DATE-TIME:20230629T113000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1011@cern.ch
DESCRIPTION:Speakers: Carmela Cappelli (University of Naples Federico II)
\nFor the analysis of ordered categorical data\, CUB modelling approach en
tails the estimation of two main structural latent components of the ratin
g process: feeling and uncertainty\, parameterized within a two-component
mixture of Binomial and uniform distributions: see Piccolo and Simone 2019
for an overview. Featuring parameters can be possibly linked to subject c
ovariates to determine twofold response patterns and they can be promptly
estimated using the EM algorithm (as implemented in the R package ‘CUB
’ available on CRAN).\nThe contribution aims at presenting how change po
int detection of temporal series of estimated feeling and uncertainty can
be pursued to identify if and to what extent Italian people modified their
perception and judgments of price levels from 1994 to 2019. To this goal\
, we resort to the framework of Atheoretical Regression Trees (ART\, Cappe
lli et al. 2008) considering the series of monthly response distributions
to questions:\n1-(Judgments): How do you think the price level changed ove
r the previous 12 months? \n2-(Expectations): How do you think the price l
evel will change over the next 12 months?\nissued by the Italian National
Statistical Institute (ISTAT) within the consumers’ confidence survey. R
esponses are collected over a scale with m=5 categories (1 =`fall '\, 2 =
`stay about the same'\, 3 = `rise slightly'\, 4 = `rise moderately'\, 5 =
`rise a lot').\nPreliminary results indicate that ART is effective in part
itioning the series into sub-intervals characterized by different levels o
f the estimated model parameters\, allowing to study and compare over time
\, the change points of both feeling and uncertainty. It’s worth noticin
g that the model parameters refer to two different aspects of the responde
nts’ perception and judgment of price level\, thus the study of their ch
ange points may reveal that they show different number and location of bre
ak dates providing a further and valuable insight into the two components
of respondents’ answers.\nPerformances of ART are also discussed compa
ratively with those of other techniques for structural change point detect
ion\, in particular with respect to Bai and Perron’s procedure as ART mi
mics this procedure.\nKeywords: price expectation\; price judgment\; Atheo
retical Regression Trees\; CUB model\; change point detection\n\nReference
s:\nC. Cappelli\, R. N. Penny\, W. S. Rea\, M. Reale (2008). Detecting mul
tiple mean breaks at unknown points in official time series\, MATHEMATICS
AND COMPUTERS IN SIMULATION\, Volume 78\, Issues 2–3\, Pages 351-356\, I
SSN 0378-4754.\nD. Piccolo\, R. Simone (2019). The class of CUB models: st
atistical foundations\, inferential issues and empirical evidence. STATIST
ICAL METHOD AND APPLICATIONS\, Volume 28\, pages 389-435.\n\nhttps://indic
o.unina.it/event/67/contributions/1011/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1011/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Mixture Models for Repeatedly Measured Survey Data
DTSTART;VALUE=DATE-TIME:20230629T090000Z
DTEND;VALUE=DATE-TIME:20230629T094000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1019@cern.ch
DESCRIPTION:Speakers: Ralitza Gueorguieva (Yale University)\nSurvey data i
tems are commonly collected on a Likert scale and may have an additional
“don’t know” category. It is also typical to have questions that are
not applicable to some individuals or to observe floor or ceiling effects
on ordinal or interval responses. These situations necessitate the use of
mixture models to properly account for the structure of the data. The mod
el formulation also needs to account for correlations among repeated measu
res within individual. We present a couple of mixture models with random e
ffects for such situations. In particular\, we use logistic sub-models to
handle “don’t know”\, inapplicable or floor effects and appropriate
generalized linear sub-models for the remaining data. Correlated random ef
fects link the sub-models together. For illustration we use data from toba
cco surveys. Maximum likelihood estimation methods are used for model fitt
ing and inference. The software implementation is using PROC NLMIXED in SA
S. Simulation studies evaluate bias and efficiency of the parameter estima
tes.\n\nhttps://indico.unina.it/event/67/contributions/1019/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1019/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Assessing replication success via skeptical mixture priors
DTSTART;VALUE=DATE-TIME:20230629T083000Z
DTEND;VALUE=DATE-TIME:20230629T085000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1014@cern.ch
DESCRIPTION:Speakers: Leonardo Egidi (Università degli Studi di Trieste)\
nThere is a growing interest in the analysis of replication studies of ori
ginal findings across many disciplines. When testing a hypothesis for an e
ffect size\, two Bayesian approaches stand out for their principled use of
the Bayes factor (BF)\, namely the replication BF (Verhagen and Wakenmale
rs\, 2014) and the skeptical BF (Pawel and Held\, 2022). In both cases rep
lication data are used to compare an "advocacy" prior against a benchmark.
For the replication BF\, the latter is the standard point null hypothesis
of no effect while for the skeptical BF it represents the prior of somebo
dy who is unconvinced by the original findings. We propose a novel skeptic
al mixture prior which incorporates skepticism and limits prior-data confl
ict. We support our proposal with theoretical results on consistency of th
e resulting BF\, we illustrate its features on an extended example\, and w
e apply it to case studies from the Social Sciences Replication Project.\n
\n**Keywords**: Bayes factor\, consistency\, prior-data conflict\,\nreplic
ation studies.\n\n**References:**\n\nPawel\, S.\, & Held\, L. (2022). The
sceptical Bayes factor for the assessment of replication success. *Journal
of the Royal Statistical Society Series B: Statistical Methodology*\, 84(
3)\, 879-911.\n\nVerhagen\, J.\, & Wagenmakers\, E. J. (2014). Bayesian te
sts to quantify the result of a replication attempt. *Journal of Experimen
tal Psychology: General*\, 143(4)\, 1457.\n\nhttps://indico.unina.it/event
/67/contributions/1014/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1014/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian networks for complementing and building gender equality c
omposite indicators
DTSTART;VALUE=DATE-TIME:20230628T150000Z
DTEND;VALUE=DATE-TIME:20230628T152000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1009@cern.ch
DESCRIPTION:Speakers: Paola Vicard (University Roma Tre)\nComposite indica
tors are a common choice for synthesizing complex phenomena. Over the year
s\, they have grown in popularity and are now applied in many social and e
nvironmental sciences. Among others\, a subject of increasing interest is
gender equality analysis. Gender composite indicators\, even if easy to re
ad\, may provide a limited picture of the problem. Here we discuss the pot
entiality of Bayesian networks (BNs) to complement and build composite ind
icators. BNs are powerful tools for explaining the complex association str
ucture in the dataset and developing scenarios to orient policy-making. He
re we propose to use BNs to model the association structure among the gend
er equality index\, its ingredient variables and other context socio-econo
mic variables. In such a way the synergy between composite indicator and B
N gives rise to both a monitoring tool for the gender equality gap status
and a proactive inferential machine for proposing policies to reduce inequ
ality. BNs can be also used to build the gender equality index\, and\, in
general\, any composite indicator. Specifically\, we focus attention on an
extension of BNs\, namely Object-Oriented Bayesian networks (OOBNs). The
modularity of the OOBN ensures a computational logic that is consistent wi
th composite indicators\, while also providing additional information abou
t the relational structure of variables. An example is carried out on Ital
ian province-level data.\n\n**Keywords:** composite indicator\, gender equ
ality\, multivariate dependencies\, Object oriented Bayesian networks\n\n*
*References:**\n\nCowell\, R. G.\, Dawid\, A. P.\, Lauritzen\, S. L.\, and
Spiegelhalter\, D. J. (1999). Probabilistic Networks and Expert Systems.
Springer Verlag\, New York\n\nMusella\, F.\, Vicard\, P. (2015). Object-or
iented Bayesian networks for complex quality management problems. Quality
& Quantity\, 49\, 115–133\n\nhttps://indico.unina.it/event/67/contributi
ons/1009/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1009/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian MANOVA for the combined evaluation of handwriting evidenc
e
DTSTART;VALUE=DATE-TIME:20230628T132000Z
DTEND;VALUE=DATE-TIME:20230628T134000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1018@cern.ch
DESCRIPTION:Speakers: Lampis Tzai ()\nForensic science is a broad field th
at uses scientific principles and technical methods to help with the evalu
ation of evidence in legal proceedings of criminal\, civil\, or administra
tive nature. Forensic scientists examine recovered traces that can be give
n by glass fragments\, fingerprints\, body fluids\, textile fibers\,\ndigi
tal device data and handwriting. The handwriting examination is a well-kno
wn field of analysis. Consider a case involving a handwritten document of
unknown origin. Handwritten features extracted from this questioned docume
nt will be compared to those extracted from a document written by a person
that is suspected of being at the origin of the anonymous document. The p
ropositions of interest are the following:\n\n$H_p$: the suspect is the au
thor of the manuscript\;\n\n$H_d$: the suspect is not the author of the ma
nuscript.\n\nHandwriting individualization is still largely dependent on t
he experience of the document examiner\, though studies have been conducte
d with the aim of supporting handwriting examiners to reduce the degree of
subjectivity of their expertise. Marquis et al. (2005) proposed to increa
se the degree of objectivity of handwriting analyses by implementing eleme
nts of Fourier analysis in order to describe the contour shape of loops of
characters. Specifically\, the characters that contain loops can be descr
ibed by means of Fourier descriptors\, which can be used to characterize t
he shape complexity and other geometric attributes. The\nanalyses conducte
d showed that these features have a good discriminating power.\nWith the a
im of implementing the use of these handwriting features for handwriting i
dentification\, Bozza et al. (2008) proposed a Bayesian probabilistic appr
oach by modeling the data with multivariate Normalinverse-Wishart distribu
tion (NIW). The value of the evidence is subsequently assessed by means of
the Bayes factor\, which can be interpreted as a measure of the strength
of support provided by the evidence in favor of the hypothesis $H_p$ again
st the hypothesis $H_d$. This approach was accomplished to take into accou
nt the correlation between variables\, the variability between-writers and
within-writer variability. However\, the above model is implemented separ
ately for each different type of handwritten character. This can be proble
matic because it can lead to a different conclusion depending on the type
of character that is retained.\n\nIn this research\, it is proposed the im
plementation of a Bayesian Multivariate Analysis of Variance (MANOVA) via
using the loop characters as predictors. The indicator of the loop charact
er is transformed into a dummy variable (corner-point representation)\, so
that it is possible to model variables describing\nthe handwriting charac
ters jointly taking into account the variability between characters\, the
variability between-writers for every character and within-writer variabil
ity. The Bayesian MANOVA is compared with the two-level random effect mode
l (NIW) proposed by Bozza et al. (2008)\, that is implemented by modelling
\nall characters jointly or separately. Three different methods for estima
ting the marginal likelihoods are used\; the Generalized Harmonic Mean\, t
he Laplace-Metropolis and the Bridge Sampling. Finally\, the performances
of the NIW and MANOVA models are compared with those of an alternative one
\, where a conjugate approach is chosen. This does not allow to model the
within and between variation separately\, but the marginals can be obtaine
d analytically.\n\nFirstly\, we estimate the Bayes factor of the two data
models for each writer to determine which model is more compatible with th
e data. Secondly\, there have been selected handwriting features originati
ng from the same writer or from different writers to evaluate the rate of
false negatives (that is cases where the BF is smaller than one for charac
ters originating from the same source) and false positives (that is cases
where the BF is greater than one for characters originating from different
sources). Finally\, the sensitivity of models is examined in two critical
aspects: the misleading background information and the choice of degrees
of freedom for the Wishart-inverse distribution that is used to model the
handwriting variability. With reference to the misleading background infor
mation\, the prior distributions were elicited by selecting writers charac
terized by either small or marked differences.\n\n**Keywords** : 1. Handwr
iting Evidence 2. Fourier Analysis 3. Multivariate Bayesian Modelling 4. B
ayes Factor 5. Sensitivity\n\n**References**\n\nBozza\, S.\, Taroni\, F.\,
Marquis\, R. & Schmittbuhl\, M. (2008)\, ‘Probabilistic evaluation of h
andwriting evidence:\nlikelihood ratio for authorship’\, Journal of the
Royal Statistical Society: Series C (Applied Statistics)\n57(3)\, 329–34
1.\n\nMarquis\, R.\, Schmittbuhl\, M.\, Mazzella\, W. D. & Taroni\, F. (20
05)\, ‘Quantification of the shape of handwritten\ncharacters: a step to
objective discrimination between writers based on the study of the capita
l character o’\,\nForensic Science International 150(1)\, 23–32.\n\nht
tps://indico.unina.it/event/67/contributions/1018/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1018/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Variable selection via ranking in generalized linear models
DTSTART;VALUE=DATE-TIME:20230628T140000Z
DTEND;VALUE=DATE-TIME:20230628T142000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1007@cern.ch
DESCRIPTION:Speakers: Marcella Niglio (Università degli Studi di Salerno)
\nIn many empirical domains\, the availability of ultrahigh-dimensional da
ta has led to the development of feature screening and variable selection
procedures aiming to detect the informative variables of datasets and cons
equently remove unimportant features.\nIn this context\, we propose a rank
ing-based variable selection procedure that extends the Ranking Based Vari
able Selection technique (Baranowski et al.\, 2020) to general linear regr
ession models.\nWe explore the performance of our proposal using both simu
lated and empirical data. The algorithm is compared to two competitors: i)
the Extended BIC (Chen and Chen\, 2012)\; ii) the variable selection proc
edure based on the combination of the Sure Independence Screening (Fan and
Song\, 2010) and the Elastic Net (Zou and Hastie\, 2005).\n\n**References
**\n\nBaranowski R\, Chen Y\, Fryzlewicz P (2020)\, Ranking-based variable
selection for high-dimensional data\, Statistica Sinica\, 30(3)\, 1485-15
16.\n\nChen J\, Chen Z (2012)\, Extended BIC for small-n-large-P sparse GL
M. Statistica Sinica\, 22(2)\, 555-574.\n\nFan J\, Song R (2010)\, Sure in
dependence screening in generalized linear models with NP-dimensionality.
The Annals of Statistics\, 38(6)\, 3567-3604.\n\nZou H\; Hastie T (2005)\,
Regularization and Variable Selection via the Elastic Net. Journal of the
Royal Statistical Society\, Series B\, 67(2)\, 301-320.\n\nhttps://indico
.unina.it/event/67/contributions/1007/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1007/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Objective Shrinkage Priors Via Imaginary Data
DTSTART;VALUE=DATE-TIME:20230628T134000Z
DTEND;VALUE=DATE-TIME:20230628T140000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1001@cern.ch
DESCRIPTION:Speakers: Dimitris Fouskakis (National Technical University of
Athens)\nIn this work\, focus is given in the Bayesian variable selection
problem for high-dimensional linear regression problems. The use of shrin
kage priors\, when the number n of available observations is less than the
number p of explanatory variables\, is a well-established method\, which
shares great theoretical and empirical properties. By using imaginary data
and shrinkage priors as baseline priors\, under the Power-Expected-Poster
ior (PEP) prior methodology\, objective shrinkage priors are being created
. In addition\, we explore the idea of augmenting the imaginary design mat
rix in order to make it with orthogonal columns and thus to produce indepe
ndent PEP-shrinkage priors\, based on default baseline priors. Under this
setup\, properly chosen hyperpriors are placed on the power parameters of
the PEP methodology\, in order to produced mixtures of independent priors
suitable for the variable selection problem when n << p. This second appro
ach provides us with algorithmically flexibility and less time-consuming p
rocedures. We check the theoretical properties of our proposed methods and
we explore their behavior via simulated studies.\n\nKeywords: Bayesian Va
riable Selection\; Imaginary Data\; Objective Priors\; Shrinkage Priors\n\
nhttps://indico.unina.it/event/67/contributions/1001/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1001/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Power Prior's Weight Parameter elicitation via Bayes Factor-Calibr
ated p-values
DTSTART;VALUE=DATE-TIME:20230629T103000Z
DTEND;VALUE=DATE-TIME:20230629T105000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1017@cern.ch
DESCRIPTION:Speakers: Roberto Macrì Demartino (University of Padova)\nIn
recent times\, the integration of historical data in the design and analys
is of new clinical trials has gained considerable attention\, owing to eth
ical reasons and difficulties encountered in recruiting patients. In the B
ayesian framework\, the process of informative prior elicitation is widely
recognized as a complex and multifaceted undertaking\, requiring the care
ful quantification and synthesis of prior information into an appropriate
prior distribution. Hence\, there is a pressing need for developing techni
ques and methods that can facilitate synthesizing and quantifying prior in
formation more effectively and efficiently. Within this context\, the conc
ept of *power priors* ([Chen and Ibrahim\, 2000][1]) has emerged as a popu
lar approach for incorporating historical data into the prior distribution
of a treatment effect\, in a flexible and controlled manner. \nThe power
prior methodology heavily relies on the *weight parameter* $\\delta$\, ran
ging between 0 and 1\, that is a crucial factor in determining the degree
to which the historical data influences the prior distribution\, and for w
hich multiple elicitation strategies are available. A modification of the
power prior allows a hierarchical prior specification by taking $\\delta$
as a random quantity\n\n$\\hspace{4.5cm}\n\\pi\\left({\\theta}\, \\delta \
\mid D_0\\right) \\propto L\\left({\\theta} \\mid D_0\\right)^\\delta \\pi
_0({\\theta}) \\pi_0(\\delta)\,\n$\n\nwhere $D_0$ is an historical dataset
with corresponding likelihood $L(\\theta \\mid D_0)$\, $\\pi_0({\\theta})
$ and $\\pi_0(\\delta)$ are the initial priors for $\\theta$ and $\\delta$
\, respectively. Furthermore\, a significant benefit of incorporating a no
rmalizing factor in the power prior methodology is its adherence to the li
kelihood principle\, as demonstrated by the joint normalized power prior\n
\n\n$\\hspace{4.5cm}\n \\pi\\left({\\theta}\, \\delta \\mid D_0\\right)
= \\frac{L\\left({\\theta} \\mid D_0\\right)^\\delta \\pi_0({\\theta}) \\
pi_0(\\delta)}{\\int_{\\Theta} L\\left({\\theta} \\mid D_0\\right)^\\delta
\\pi_0({\\theta}) d {\\theta}}.\n$\n\nConsequently\, in a fully Bayesian
approach\, the ability to effectively elicit an appropriate initial prior
distribution for the weight parameter $\\delta$ is a crucial step. As far
as we know from reviewing the existing literature\, a comprehensive justif
ication underlying the choice of a Beta distribution with fixed hyper-para
meters\, that is an usual choice for this framework\, is pretty vague. \n\
nThe Bayes factor (BF) constitutes a valuable statistical tool for model c
omparison\; however\, we explore the use of the Bayes Factor to discrimina
te between competing models that incorporate distinct initial Beta prior d
istributions for the weight parameter by exploiting some BF $p$-value cali
bration techniques ([Garcia-Donato and Chen\, 2005][2]). \nThis would enab
le the selection of candidate models based on a more accurate and reliable
assessment of the available evidence\, thereby enhancing the validity and
robustness of statistical inference.\n\n**Keywords:** Beta distribution\,
Clinical trial\, Historical information\, Robust selection.\n\n\n**Refere
nces**\n\n 1. Chen\, M.-H. and Ibrahim\, J. G. (2000). Power prior distrib
utions for regression models. Statistical Science\, 15(1):46 – 60.\n\n 2
. Garcia-Donato\, G. and Chen\, M.-H. (2005). Calibrating Bayes factor\n
under prior predictive distributions. Statistica Sinica\, 15(2):359–38
0.\n\n [1]: https://www.jstor.org/stable/2676676\n [2]: https://www.jsto
r.org/stable/24307360\n\nhttps://indico.unina.it/event/67/contributions/10
17/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1017/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Regularized Joint Mixture Models
DTSTART;VALUE=DATE-TIME:20230629T130000Z
DTEND;VALUE=DATE-TIME:20230629T132000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1021@cern.ch
DESCRIPTION:Speakers: Konstantinos Perrakis (Department of Mathematical Sc
iences\, Durham University)\nRegularized regression models are well studie
d and\, under appropriate conditions\, offer fast and statistically interp
retable results. However\, large data in many applications are heterogeneo
us in the sense of harboring distributional differences between latent gro
ups. Then\, the assumption that the conditional distribution of response Y
given features X is the same for all samples may not hold. Furthermore\,
in scientific applications\, the covariance structure of the features may
contain important signals and its learning is also affected by latent grou
p structure. We propose a class of mixture models for paired data (X\, Y)
that couples together the distribution of X (using sparse graphical models
) and the conditional Y | X (using sparse regression models). The regressi
on and graphical models are specific to the latent groups and model parame
ters are estimated jointly (hence the name "regularized joint mixtures").
This allows signals in either or both of the feature distribution and regr
ession model to inform learning of latent structure and provides automatic
control of confounding by such structure. Estimation is handled via an ex
pectation-maximization algorithm\, whose convergence is established theore
tically. We illustrate the key ideas via empirical examples. An R package
is available at https://github.com/k-perrakis/regjmix.\n\n*Keywords:* dist
ribution shifts\, heterogeneous data\, joint learning\, latent groups\, mi
xture\nmodels\, sparse regression\n\nhttps://indico.unina.it/event/67/cont
ributions/1021/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1021/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian effect measures for a location scale model
DTSTART;VALUE=DATE-TIME:20230629T134000Z
DTEND;VALUE=DATE-TIME:20230629T140000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1016@cern.ch
DESCRIPTION:Speakers: Claudia Tarantola (University of Pavia)\nWe consider
a Bayesian approach for the analysis of rating data when a scaling compon
ent is taken into account\, thus incorporating a specific form of heterosk
edasticity. Our approach includes model-based probability effect measure
s that enable comparisons of distributions among multiple groups. These ef
fect measures are adjusted for explanatory variables that have an impact o
n both the location and scale components. To estimate the parameters of ou
r fitted model and derive the associated effect measures\, we employ Marko
v Chain Monte Carlo techniques. Through an analysis of students' evaluatio
ns of a university curriculum counselor service\, we assess the performanc
e of our method and highlight its valuable support in the decision-making
process. Our findings demonstrate the effectiveness of our approach and em
phasize its ability to enhance decision-making processes by providing valu
able insights and support to stakeholders involved.\n\nhttps://indico.unin
a.it/event/67/contributions/1016/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1016/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Bayesian learning of network structures from interventional experi
mental data
DTSTART;VALUE=DATE-TIME:20230630T075000Z
DTEND;VALUE=DATE-TIME:20230630T081000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1026@cern.ch
DESCRIPTION:Speakers: Stefano Peluso ()\nDirected Acyclic Graphs (DAGs) pr
ovide an effective framework for learning causal relationships among varia
bles given multivariate observations. Under pure observational data\, DAGs
encoding the same conditional independencies cannot be distinguished and
are collected into Markov equivalence classes. In many contexts however\,
observational measurements are supplemented by interventional data that im
prove DAG identifiability and enhance causal effect estimation. We propose
a Bayesian framework for multivariate data partially generated after stoc
hastic interventions. To this end\, we introduce an effective prior elicit
ation procedure leading to a closed-form expression for the DAG marginal l
ikelihood and guaranteeing score equivalence among DAGs that are Markov eq
uivalent post intervention. Under the Gaussian setting we show\, in terms
of posterior ratio consistency\, that the true network will be asymptotica
lly recovered\, regardless of the specific distribution of the intervened
variables and of the relative asymptotic dominance between observational a
nd interventional measurements. We validate our theoretical results in sim
ulation and we implement on both synthetic and biological protein expressi
on data a Markov chain Monte Carlo sampler for posterior inference on the
space of DAGs.\n\nhttps://indico.unina.it/event/67/contributions/1026/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1026/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Integrating model-based clustering and graphical models to explore
the relationship with the digital self-image in (pre)adolescents
DTSTART;VALUE=DATE-TIME:20230630T081000Z
DTEND;VALUE=DATE-TIME:20230630T083000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1015@cern.ch
DESCRIPTION:Speakers: Chiara Brombin (CUSSB\, Faculty of Psychology\, Vita
-Salute San Raffaele University)\nDigital revolution has dramatically chan
ged not only the way people interact but also the relationship with the se
lf-image. Increased data availability and computational power have signifi
cantly improved algorithms for facial feature detection which have been al
so successfully applied to develop face filter apps enhancing and “beaut
ifying” self-portraits.\nPotential of these filters in altering facial a
ppearance has raised concerns in parents\, educators and health profession
als as they promote unrealistic beauty standards increasing discrepancy be
tween real and digital self. Actually taking\, sharing and viewing edited
selfies may have detrimental effects especially on younger users in a deve
lopmental phase where they are already facing significant identity constru
ction processes\, possibly giving rise to appearance-related cyberbullying
. \nTo investigate selfie-sharing/editing behaviour in (pre)adolescents\,
their relationship with digital self-image\, problematic use of social net
work and possible internalizing symptoms an online questionnaire\, includi
ng both validated and ad-hoc realized scales\, has been developed. When ex
amining the digital-self image\, here the attention is narrowed to the fac
e only\, the protagonist of real and virtual interactions\, and not to the
whole body. \nIn this setting\, graphical models represent an appealing t
ool to model dependence structure between collected variables. To properly
analyze collected data\, the procedure should account for the fact that (
i) data from psychological questionnaires are usually measured on discrete
/ordinal levels thus violating the normality assumption and (ii) measured
behaviors are rarely homogeneous and this heterogeneity should be properly
modelled to obtain unbiased results. \nTo tackle these issues\, an approa
ch integrating model-based clustering and graphical models (Fop et al.\, 2
019)\, has been applied to copula transformed data collected on a sample o
f 229 middle school (pre)adolescents which took part to the online survey.
A two-clusters solution was selected as best based on BIC criterion: the
two clusters actually showed different covariance network and different ma
nagement of online self-image and psychological status. Participants in th
e cluster displaying a worse management of online self-image and psycholog
ical status were mainly female reporting higher use of social networks. To
better examine the relationships among variables within each cluster\, pa
rtial correlation networks were estimated separately for the two clusters
and compared using both global and local network statistics and inferenti
al procedure for network comparison.\nAlthough graphical models have been
widely used to model psychological phenomena as complex networks\, the app
lication to selfie behavior is original. Moreover\, identifying clusters w
ithin a graphical model framework has important practical implications suc
h as (i) aiding in the development of tailored training programs suited fo
r improving digital wellbeing in younger users and (ii) uncovering new da
ta-driven relationships among constructs thus generating new hypothesis to
test in successive studies.\n\nReference\nFop\, M.\, Murphy\, T.B. and Sc
rucca\, L.\, 2019. Model-based clustering with sparse covariance matrices.
Statistics and Computing\, 29(4)\, pp.791-819. \nKashihara\, J.\, Takebay
ashi\, Y.\, Kunisato\, Y. and Ito\, M. (2021). Classifying patients with d
epressive and anxiety disorders according to symptom network structures: A
Gaussian graphical mixture model-based clustering. Plos one\, 16(9)\, p.e
0256902.\n\nhttps://indico.unina.it/event/67/contributions/1015/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1015/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Network Integration with INet algorithm
DTSTART;VALUE=DATE-TIME:20230630T105000Z
DTEND;VALUE=DATE-TIME:20230630T111000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1027@cern.ch
DESCRIPTION:Speakers: Valeria Policastro ()\nNowadays\, network data integ
ration is a demanding problem and still an open challenge\, especially whe
n dealing with large datasets. When collecting several data sets and heter
ogeneous data types on a given phenomenon of interest\, the individual ana
lysis of each data set will give only a particular view of such phenomenon
. In contrast\, integrating all the data will widen and deepen the results
giving a more global view of the entire system.\n\nWe developed a novel s
tatistical method named INet algorithm\, for data integration based on wei
ghted multilayer networks. Under the assumption that the structure underne
ath the different layers has some similarity that we want to emerge in the
integrated network\, we generate a “consensus network” through an ite
rative procedure based on structure comparison\, capable of pulling out im
portant information about the phenomenon under study. The procedure tries
to preserve common higher-order structures of the original networks in the
integrated one\, i.e. neighborhood. Once obtained the consensus network\,
we compared it with the starting networks extracting “specific networks
”\, one for each layer\, containing peculiar information of the single d
ata type not present in all the others.\n\nWe tested our method on simulat
ed networks to analyze the performance of our algorithm and we analyzed vi
rus and vaccine gene co-expression networks to better understand infectiou
s diseases.\n\nhttps://indico.unina.it/event/67/contributions/1027/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1027/
END:VEVENT
BEGIN:VEVENT
SUMMARY:How to peel the network: an algorithm for weighted triad census
DTSTART;VALUE=DATE-TIME:20230630T103000Z
DTEND;VALUE=DATE-TIME:20230630T105000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1023@cern.ch
DESCRIPTION:Speakers: Roberto Rondinelli ()\nIn Network Analysis\, the int
eraction between three nodes is called a "triad" and represents the minima
l group structure that can be observed. According to the presence and the
type of the relations between three nodes\, sixteen triadic configurations
(called the isomorphism classes) are defined and their distribution is de
noted as "triad census". This kind of analysis is used in different situat
ions concerning relational data and the conventional approach is well-defi
ned for unweighted networks. As a consequence\, the information regarding
the weights is not taken into account. \nTo exploit this information in th
e triad analysis\, we propose a new algorithm denoted as "network peeling"
to count the different configurations of triads in weighted networks. The
algorithm computes the triad census over the network layers generated at
each step. The resulting matrix (with dimensions layers x isomorphism clas
ses) can be summarized through a set of descriptive measures representing
the weighted triad census.\nWith the aim to highlight the appropriateness
of our approach\, we consider some real scenarios and a simulation study\,
comparing weighted and conventional triad censuses.\n\nhttps://indico.uni
na.it/event/67/contributions/1023/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1023/
END:VEVENT
BEGIN:VEVENT
SUMMARY:φ-Divergence based Modelling of Categorical and Rank Data
DTSTART;VALUE=DATE-TIME:20230629T075000Z
DTEND;VALUE=DATE-TIME:20230629T081000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1022@cern.ch
DESCRIPTION:Speakers: Maria Kateri (RWTH Aachen University)\nStandard mode
ls for categorical and ordinal data\, such as log-linear\, association mod
els and logistic regression models for binary or ordinal responses\, as we
ll as the Mallows model for rank data are revisited and defined through st
atistical information theoretic properties in terms of the Kullback–Leib
ler (KL) divergence. In the sequel\, replacing the KL by the φ-divergence
\, which is a family of divergences including the KL as special case\, the
se models are generalized to flexible families of models. The suggested mo
dels are discussed in terms of their properties\, estimation and fit. Fina
lly\, their potential is illustrated by characteristic examples.\n\nKey-wo
rds: Cressie–Read power divergence\, distance-based probability models\,
maximum likelihood estimation\n\nhttps://indico.unina.it/event/67/contrib
utions/1022/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1022/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Fast Bayesian Variable Screening Using Correlation Thresholds
DTSTART;VALUE=DATE-TIME:20230629T081000Z
DTEND;VALUE=DATE-TIME:20230629T083000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1012@cern.ch
DESCRIPTION:Speakers: Ioannis Ntzoufras (AUEB)\nWe propose a fast Bayesian
variable selection method for Normal regression models\, using Zellner's
$g$-prior specification. The approach is based on using thresholds on Pear
son and partial correlation coefficients. Nevertheless\, the proposed meth
odology is derived using purely Bayesian arguments derived from thresholds
on Bayes factors and posterior model odds. \nThe proposed method can be u
sed to screen out the non-important covariates and reduce the model space
size. Then\, traditional\, computer-intensive\, Bayesian variable selectio
n methods can be implemented without any problem with the derived reduced
model space. \nWe focus on the g-prior where the Bayes factor computations
and the corresponding correlation thresholds are exact. Nevertheless\, th
e approach is general and can be easily extended to any prior setup. \nThe
proposed method is illustrated using simulated examples.\n\nhttps://indic
o.unina.it/event/67/contributions/1012/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1012/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Modelling ordinal data from repeated surveys
DTSTART;VALUE=DATE-TIME:20230629T073000Z
DTEND;VALUE=DATE-TIME:20230629T075000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1010@cern.ch
DESCRIPTION:Speakers: Marcella Corduas (Departmento of Political Sciences\
, University of Naples Federico II)\nBusiness and consumers survey data ar
e the basis for several indicators describing the trend of macro-economic
variables that are fundamental for monitoring the overall performance of t
he economic system. Qualitative surveys typically ask interviewees to expr
ess their perceptions or expectations about the current or future tendency
of a reference economic variable (such as inflation or industrial output)
using a trichotomous or a finer-tuned ordered scale. Surveys are carried
out at regular interval by statistical offices\, and collected data are t
raditionally published in aggregate form\, reporting the proportions of po
sitive\, neutral or negative assessments. This contribution presents an in
novative dynamic model that describes the probability distributions of ord
ered categorical variables observed over time. For this aim\, we extend th
e definition of the mixture distribution obtained from the Combination of
a Uniform and a shifted Binomial distribution (CUB model)\, introducing ti
me varying parameters. The model parameters identify the main components r
uling the respondent evaluation process: the degree of attraction towards
the object under assessment\, the uncertainty related to the answer\, and
the weight of the refuge category that is selected when a respondent is un
willing to elaborate a thoughtful judgement. We suggest to use the model t
ime-varying parameters as indicators of the diversity of respondents' opin
ions\, shifting from an optimistic to a pessimistic state as the surroundi
ng conditions evolve. For illustrative purpose\, the dynamic CUB model is
applied to the consumers' perception and expectations of inflation in Ita
ly to investigate: a) the effect of the COVID pandemic on the respondents
’ perceptions\; b) the impact of the respondents' income level on expect
ations.\n\n**Keywords:** ordinal data\; CUB model\; consumers' perceptions
\; consumers' expectations\n\n**References**\n\nCorduas\, M.: A dynamic mo
del for ordinal time series: An application to consumers' perceptions of
inflation. In: *Statistical Learning and Modeling in Data Analysis*\, Bal
zano\, S.\, Porzio\, G.C.\, Salvatore\, R.\, Vistocco\, D.\, Vichi\, M.
(Eds.)\, Cham: Springer 2019\, (pp. 37-45).\n\nPiccolo\, D.\, Simone\, R
. The class of cub models: statistical foundations\, inferential issues an
d empirical evidence\, (with discussion and rejoinder). *Stat. Meth.\\& Ap
pl.* 2019\, 28\, 389-435.\n\nhttps://indico.unina.it/event/67/contribution
s/1010/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1010/
END:VEVENT
BEGIN:VEVENT
SUMMARY:A model-robust subsampling approach in presence of outliers
DTSTART;VALUE=DATE-TIME:20230628T154000Z
DTEND;VALUE=DATE-TIME:20230628T160000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1006@cern.ch
DESCRIPTION:Speakers: Laura Deldossi ()\nIn the era of big data\, several
sampling approaches are proposed to reduce costs (and time) and to help in
informed decision making. Some of these proposals (Drovandi et al.\, 2017
\; Wang et al.\, 2019\; Deldossi and Tommasi (2022) among others) are insp
ired to Optimal Experimental Design and require the specification of a mod
el for the big dataset. \nThis model assumption\, as well as the possible
presence of outliers in the big dataset represent a limitation for the mos
t commonly applied subsampling criterions.\nDeldossi et al. (2023) introdu
ced non-informative and informative exchange algorithms to select “nearl
y” D-optimal subsets without outliers in a linear regression model. \n\n
In this study\, we extend their proposal to account for model uncertainty.
More precisely\, we propose a model robust approach where a set of candid
ate models is considered\; the optimal subset is obtained by merging the s
ubsamples that would be selected by applying the approach of Deldossi et a
l. (2023) if each model was considered as the true generating process. \nT
he approach is applied in a simulation study and some comparisons with oth
er subsampling procedures are provided.\n\nKey-words: Active learning\, D-
optimality\, Subsampling \n\nReferences\n\nDeldossi\, L.\, Tommasi C. (202
2) Optimal design subsampling from Big Datasets. Journal of Quality Techn
ology 54(1): 93–101\n\nDeldossi\, L.\, Pesce\, E.\, Tommasi\, C. (2023)
Accounting for outliers in optimal subsampling methods\, Statistical Paper
s\, https://doi.org/10.1007/s00362-023-01422-3.\n\nDrovandi CC\, Holmes CC
\, McGree JM\, Mengersen K\, Richardson S\, Ryan EG (2017) Principles of e
xperimental design for big data analysis. Statistical Sciences 32(3): 385
–404\n\nWang H\, Yang M\, Stufken J (2019) Information-based optimal sub
data selection for Big Data linear regression. Journal of American Statis
tical Association 114(525): 393–405\n\nhttps://indico.unina.it/event/67/
contributions/1006/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1006/
END:VEVENT
BEGIN:VEVENT
SUMMARY:A structural equation model to integrate item responses\, response
times and item positioning in students' ability assessment
DTSTART;VALUE=DATE-TIME:20230628T152000Z
DTEND;VALUE=DATE-TIME:20230628T154000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1005@cern.ch
DESCRIPTION:Speakers: Carla Galluccio (Department of Statistics\, Computer
Science\, Applications "G. Parenti")\nIn the context of students' ability
assessment\, considering collateral information in addition to item respo
nses can be helpful in increasing the accuracy of the measurement. In this
vein\, the evaluation of students' abilities via computer based-devices
has made response time data available at the item level (Wang et al.\, 201
9). Besides\, the literature (Becker et al.\, 2022) has highlighted an ite
m position effect when the same items are presented in different positions
within multiple test forms. \n\nWith the present contribution\, we contri
bute to this research line by proposing a structural equation model (SEM)
to jointly consider item responses\, response times and item positioning i
n students' ability assessment. In particular\, we assume that the respons
e process is driven by two underlying latent variables: the first latent v
ariable\, denoted by $\\Theta_i$\, represents the ability of individual $i
$ that is measured by the test items\; the second latent variable\, denote
d by $\\eta_i$\, refers to the speediness of individual $i$ to answer the
test items. \n\nWe formulate the statistical model assuming that the item
responses are directly affected by the ability $\\Theta_i$\, whereas the
response times depend both on the ability $\\Theta_i$ and on the speedines
s $\\eta_i$. Accordingly\, response accuracy tends to increase with the ab
ility level of individual $i$ while response time tends to decrease with t
he speediness and ability levels. Moreover\, we suppose that item position
ing affects both item responses and response time. Under this setting\, th
e correlation between $\\Theta_i$ and $\\eta_i$ is modelled through the cr
oss-relation function that models the relationships between $\\Theta_i$ an
d the observed response times.\n\nThe empirical application of the propose
d model was carried out on first-year Psychology students at the Universit
y of Naples Federico II\, attending the introductory Statistics course. Th
e test administered was composed of 30 multi-choice questions developed ac
cording to three of the five Dublin descriptors: Knowledge (10 items)\, Ap
plication (10 items) and Judgement (10 items). For each question\, student
s' answers were coded as correct (2 credits)\, partially correct (1 credit
) and wrong (0 credits). Data were collected through Moodle platform\, whi
ch also provided the response time.\n\nReferences\n\nBecker\, B.\, Van Rij
n\, P.\, Molenaar\, D.\, and Debeer\, D. (2022). Item order and speedednes
s: Implications for test fairness in higher educational high-stakes testin
g. Assessment & Evaluation in Higher Education\, 47(7):1030–1042.\n\nWan
g\, C.\, Weiss\, D. J.\, and Su\, S. (2019). Modeling response time and re
sponses in multidimensional health measurement. Frontiers in psychology\,
10:51.\n\nhttps://indico.unina.it/event/67/contributions/1005/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1005/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Subdata selection for big data regression based on leverage scores
DTSTART;VALUE=DATE-TIME:20230628T130000Z
DTEND;VALUE=DATE-TIME:20230628T132000Z
DTSTAMP;VALUE=DATE-TIME:20240619T160058Z
UID:indico-contribution-67-1004@cern.ch
DESCRIPTION:Speakers: Vasilis Chasiotis (Athens University of Economics an
d Business\, Department of Statistics)\nData continues to become more abun
dant\, and so the datasets that contain it. Even though big datasets can p
resent insights and opportunities\, they can pose significant challenges w
hen it comes to statistical analysis. One of the biggest challenges\, requ
ired to process and analyze large datasets\, is the computational resource
s. Regression can be problematic in case of big datasets\, due to the huge
volumes of data. A standard approach is subsampling that aims at obtainin
g the most informative portion of the big data. We consider an approach ba
sed on leverages scores\, already existing in the current literature for
the selection of subdata for linear model discrimination. However\, we hig
hlight its importance on the selection of data points that are the most in
formative for estimating unknown parameters. We conclude that the approach
based on leverage scores improves existing approaches\, providing simulat
ion experiments as well as a real data application.\n\nhttps://indico.unin
a.it/event/67/contributions/1004/
LOCATION:Department of Political Sciences Aula Spinelli
URL:https://indico.unina.it/event/67/contributions/1004/
END:VEVENT
END:VCALENDAR