[ABE-L] Seminário PIPGEs com Jochen Einbeck: 13/4 (segunda) às 14hs no CINA-UFSCar

[Rafael Izbicki]

Caros,

Excepcionalmente na *segunda-feira*, 13 de abril às 14hs, teremos o
seguinte seminário do Programa Interinstitucional de Pós-Graduação em
Estatística (PIPGEs) da UFSCar-USP.

*Nome:* Jochen Einbeck (Durham University, UK)

*Título:* Robust ratio estimators for location parameters of count
distributions

*Resumo:* Classical parameter estimation for Poisson-distributed data
relies mainly on Maximum Likelihood Estimation (MLE), which assumes that
the data perfectly adhere to the model assumptions including
equidispersion. However, real-world datasets often suffer from structural
discrepancies that violate these assumptions, leading to biased estimates
and inflated mean squared errors. This study derives two robust
estimators—the Single Ratio (SR) and the Double Ratio (DR) estimators,
which are based on the information in the data close to the mode.
Consistency of the estimators is established analytically. Using a Monte
Carlo simulation, we evaluate the performance of these estimators across
four critical scenarios: outliers, right truncation, zero inflation, and
zero truncation. In all these scenarios, the proposed estimators showed
consistent behaviour, and in most scenarios, superior performance to the
MLE, especially for larger sample sizes. Real data-examples illustrate the
proposed estimators. Extension to other count distributions, as well as the
development of a GLM-type regression framework for this class of
estimators, are briefly considered. (joint work with Dr Sehar Saleem).

*Local:* CINA-UFSCar (São Carlos)

Todos são bem-vindos!


[image: Seminario.png]

--
Rafael Izbicki
Associate Professor | Vice Director of Graduate Studies
Department of Statistics
Federal University of São Carlos (UFSCar)
https://rafaelizbicki.com/
https://small-research.github.io/website
-------------- Próxima Parte ----------
Um anexo em HTML foi limpo...
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20260401/c3cf99a4/attachment-0001.htm>
-------------- Próxima Parte ----------
Um anexo não-texto foi limpo...
Nome: Seminario.png
Tipo: image/png
Tamanho: 311462 bytes
Descrição: não disponível
URL: <http://lists.ime.usp.br/pipermail/abe/attachments/20260401/c3cf99a4/attachment-0001.png>