NEWS
{Jacobo de
Uña-Álvarez web site}
[Research]
[21/01/15] Micha Mandel, from the Stats Department at the Hebrew
University of Jerusalem, is visiting Vigo next February
Micha
Mandel, from the Department of Statistics at the Hebrew University of
Jerusalem, is visiting the University of Vigo next February 9-13, 2015. His
main research interest is in applied statistics, especially in problems
related to medical and biological data. Dr. Micha
Mandel (Ph.D. 2004) has contributed several papers in top JCR journals like Biometrika, Biostatistics, Statistics
in Medicine, JASA, or Annals of
Applied Statistics. Dr. Mandel will give an open conference on Tuesday
February 10 at 13:00 hs. at
the Grade Room of the Faculty of Economics, with title ‘Estimating Time to
Disease Progression Using Panel Data, with Application to Multiple Sclerosis’.
Parts of this talk are joint work of Dr. Mandel with researchers Rebecca Betensky, Francois Mercier, Susan Gauthier, Howard Weiner,
Charles Guttmann, Ben Eckert, and Peter Chin.
[15/01/15] A
substitute for the Aalen-Johansen estimator in non-Markov multi-state models
has been proposed (to appear in Biometrics,
the journal of the IBS)
In
cooperation with Luis Meira-Machado, from the
University of Minho, Jacobo de Uña-Álvarez
has developed a new nonparametric method to estimate transition probabilities
in a general, non-Markov multi-state model. This new method is a suitable
substitute for the Aalen-Johansen estimator when the process does not satisfy
the Markov condition. The method, based on the idea of computing survival in
specific subsamples, overcomes some of the limitations of the original
non-Markov estimator studied in Meira-Machado’s Ph.D.
thesis (published in Lifetime
Data Analysis, back in 2006). The contribution has been presented last
December during the 7th International Conference of the ERCIM WG on Computational
and Methodological Statistics held in Pisa, and it will appear in Biometrics, the official journal of the
International Biometric Society, in the next few weeks.
Multi-state
models have been widely used in biomedicine to investigate the progression of
patients undergoing a given illness or surgery; usually, the states represent
the occurrence of an event which may be related to survival prognosis, such as
complications after surgery, recurrences, or non-fatal episodes. In these
models, the estimation of the transition probabilities is of particular
interest, since they allow for long-term predictions for the diseased
individuals.
[14/12/18] Irene
Castro-Conde defends her Ph.D. on the Sequential
Goodness-of-Fit multiple testing procedure
On December
18th, 2014, Irene Castro-Conde defended
her Ph.D. thesis entitled ‘Advances in Multiple Hypothesis Testing: the
Sequential Goodness-of-Fit Procedure Revisited and Expanded’ in the Grade Room
at the Faculty of Economics in Vigo, achieving the maximum score of ‘Sobresaliente Cum Laude’. In the thesis, Irene revisited
the SGoF method, originally introduced in 2009 by
local geneticists Antonio Carvajal-Rodríguez and
Emilio Rolán-Álvarez in cooperation with statistician
Jacobo de Uña-Álvarez. The
thesis contributes a number of relevant results on SGoF,
among them: the evaluation of false discovery rate and power of the
Beta-Binomial SGoF in simulated scenarios, the
definition and application of adjusted p-values for SGoF,
the extension of the method for discrete p-values and Bayesian scenarios, and
the development of a user-friendly R package. These contributions have already
appeared (or will appear soon) in good statistical journals, like Biometrical Journal, The R Journal, and Computational Statistics. Irene’s thesis was supervised by Jacobo de Uña-Álvarez and, during
a three months scondment, by Sebastian Doehler (at Darmstadt University of Applied Sciences) too.
The thesis jury was constituted by the well-known statisticians Daniel Yekutieli (Tel Aviv University), Frank Bretz
(Novartis at Bassel and Hannover University), and Conchi Ausín (Carlos III
University in Madrid).
Irene’s did a
good job as a predoc student at the Statistics and
Operations Research doctorate program, finishing her thesis in a relatively
short time (less than two years). Along this time, she has participated in a
number of local and international seminars and conferences, like the Multiple
Comparison Procedure meeting held in Southampton in 2013 (MCP2013) or the 7th
International Conference of the ERCIM WG on Computational and Methodological
Statistics (Pisa, 2014).
SGoF
method performs multiple hypothesis testing by comparing the observed and the
expected proportions of p-values falling below the given significance
threshold. It is related to the well-known notion of higher criticism or second
level significance testing introduced by famous mathematician Tukey in 1976 when teaching Statistics at Princeton
University. SGoF uses this notion to identify the
false null hypotheses in such a way that a large power is obtained under a
reasonable false discovery rate. Since 2009, its year of publication in BMC Bioinformatics,
SGoF has been found a truly practical tool in applied
research; indeed Web of Science reports
presently up to 50 cites for this paper, including journals like Journal of Proteomics, NMR in Biomedicine, Neuroimage, Neurochemistry International or Evolutionary Applications.