[Forum SIS] DEI - Università di Catania: AVVISO DI SEMINARI

[antonio punzo]

********************************************************************************
AVVISO DI SEMINARI
********************************************************************************

Il Dipartimento di Economia e Impresa dell’Università di Catania ha
organizzato i seminari:



*Component-based Path Modeling*

*Pasquale Dolce*

*Dipartimento di Economia e Statistica*

*Università degli Studi di Napoli “Federico II”*





Mercoledì 1 Luglio dalle ore 15:30

In Aula 6



Università di Catania

Palazzo delle Scienze, Corso Italia, 55, Catania





*Abstract*:



Partial Least Squares Path Modeling (PLS-PM) is a component-based
estimation approach for Structural Equation Modeling (SEM), that allows us
to analyze relationships among blocks of observed variables, called
manifest variables (MV), where each block is represented by a so-called
latent variable (LV). PLS-PM has some inconsistencies in terms of coherence
with the direction of the relationships specified in the structural model.
The estimation process analyses interdependence among blocks and misses to
distinguish between dependent and explanatory LVs. In this regard, it has
been proposed a method, called Non-Symmetrical Component-based Path
Modeling (NSC-PM), based on the optimization of a redundancy-related
criterion in a multi-block framework. Another issue that might arise in
PLS-PM is related to the case where there is more than a single coefficient
describing the relationship between variables. A new method, called
Quantile Composite-based Path Modelling (QC-PM), introduces the Quantile
regression (QR) in the classical PLS-PM algorithm, in order to enhance
PLS-PM potentialities when we wish to distinguish regressor effects on
different parts of the dependent variable distributions.





I colleghi interessati sono cordialmente invitati a partecipare.





*Three-step estimation approach in the context of latent Markov modeling
with covariates*

*Roberto Di Mari*

*University of RomeTor Vergata and Tilburg University*





Mercoledì 1 Luglio dalle ore 16:45

In Aula 6



Università di Catania

Palazzo delle Scienze, Corso Italia, 55, Catania





*Abstract*:



The estimation of latent Markov (LM) models with covariates is commonly
carried out via Maximum Likelihood. One-step Maximum Likelihood estimation
in the context of LM models has two main drawbacks. First, the inclusion of
covariates might alter the choice of the number of classes. Second, the
estimation routine becomes unfeasible when the algorithm deals with many
responses, a large number of covariates and time points. In addition,
researchers might be interested in building a classification model at a
first stage of the analysis. The present paper wants to address the
question of whether, in a LM context, it is possible to formulate an
alternative to one-step estimation, which might work well and facilitate
the use of many responses and covariates. The issues of time structure in
the data in the first step and classification error in the second step are
tackled. In order to motivate the use of the methodology we present an
application on household portfolio choice. Moreover, simulation study
results will also be presented.





I colleghi interessati sono cordialmente invitati a partecipare.






Antonio Punzo
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20150624/4b4358d9/attachment-0001.html>