Most classical statistical procedures are based on the assumption that  the model that generates the data is known. For this reason, these methods are very sensitive to small deviations from the model, such as the presence of atypical observations. The aim of robust statistical procedures is to allow inference that is valid even when the model is true only approximatelywhile , at thesame time,is highly efficient under the model.

Classically, the models most usually used are parametric ones, in which it is assumed that the observations belong to a known parametric family. This is a strongassumption, since the proposed parametric model may be incorrect. Moreover, statistical methods developed for a particular model may lead to wrong conclusions if applied to a slightly different one. These problems led to the development of semiparametric methods for data analysis, besides robust statistical procedures.

The Instituto de Cálculo counts with a group of professors, researchers, teaching assistants and students that carry out their research activities in the area of statistics, especially in topics that involve robust and/or semiparametric statistical methods with applications to generalized linear models, time series , partially linear regression models, common principal components models, partially linear generalized models, tolerance regions for multivariate data, sliced regression, discriminant analysis, regression models with censored data,  regression models with missing data, additive models, functional data, isotonic regression, GARCH and ARMA models, statistics in Riemannian manifolds, robust estimation of location and covariance for multivariate data with and without  missing data.


Choque, Federico
Parada, Daniela
María Florencia Statti
Álvarez, Enrique E.
Bianco, Ana
Rodríguez, Daniela
Sued, Mariela
Szretter Noste, M. Eugenia
Valdora, Marina
Yohai, Víctor J.
Boente, Graciela