-
Sparse inference in Poisson Log-Normal model by approximating the L0-norm
Authors:
Togo Jean Yves Kioye,
Paul-Marie Grollemund,
Jocelyn Chauvet,
Pierre Druilhet,
Erwan Saint-Loubert-Bie,
Christophe Chassard
Abstract:
Variable selection methods are required in practical statistical modeling, to identify and include only the most relevant predictors, and then improving model interpretability. Such variable selection methods are typically employed in regression models, for instance in this article for the Poisson Log Normal model (PLN, Chiquet et al., 2021). This model aim to explain multivariate count data using…
▽ More
Variable selection methods are required in practical statistical modeling, to identify and include only the most relevant predictors, and then improving model interpretability. Such variable selection methods are typically employed in regression models, for instance in this article for the Poisson Log Normal model (PLN, Chiquet et al., 2021). This model aim to explain multivariate count data using dependent variables, and its utility was demonstrating in scientific fields such as ecology and agronomy. In the case of the PLN model, most recent papers focus on sparse networks inference through combination of the likelihood with a L1 -penalty on the precision matrix. In this paper, we propose to rely on a recent penalization method (SIC, O'Neill and Burke, 2023), which consists in smoothly approximating the L0-penalty, and that avoids the calibration of a tuning parameter with a cross-validation procedure. Moreover, this work focuses on the coefficient matrix of the PLN model and establishes an inference procedure ensuring effective variable selection performance, so that the resulting fitted model explaining multivariate count data using only relevant explanatory variables. Our proposal involves implementing a procedure that integrates the SIC penalization algorithm (epsilon-telescoping) and the PLN model fitting algorithm (a variational EM algorithm). To support our proposal, we provide theoretical results and insights about the penalization method, and we perform simulation studies to assess the method, which is also applied on real datasets.
△ Less
Submitted 25 March, 2024;
originally announced March 2024.
-
Ecient spatial designs for targeted regions or level set detection
Authors:
Sylvain Coly,
Pierre Druilhet,
Nourddine Azzaoui
Abstract:
Acquiring information on spatial phenomena can be costly and time-consuming. In this context, to obtain reliable global knowledge, the choice of measurement location is a crucial issue. Space-lling designs are often used to control variability uniformly across the whole space. However, in a monitoring context, it is more relevant to focus on crucial regions, especially when dealing with sensitive…
▽ More
Acquiring information on spatial phenomena can be costly and time-consuming. In this context, to obtain reliable global knowledge, the choice of measurement location is a crucial issue. Space-lling designs are often used to control variability uniformly across the whole space. However, in a monitoring context, it is more relevant to focus on crucial regions, especially when dealing with sensitive areas such as the environment, climate or public health. It is therefore important to choose a relevant optimality criterion to build models adapted to the purpose of the experiment. In this article, we propose two new optimality criteria: the rst aims to focus on areas where the response exceeds a given threshold, while the second is suitable for estimating sets of levels. We introduce several algorithms for constructing optimal designs. We also focus on cost-eective algorithms that produce non-optimal but ecient designs. For both sequential and non-sequential contexts, we compare our designs with existing ones through extensive simulation studies.
△ Less
Submitted 14 February, 2024;
originally announced February 2024.
-
Prediction model for rare events in longitudinal follow-up and resampling methods
Authors:
Pierre Druilhet,
Mathieu Berthe,
Stéphanie Léger
Abstract:
We consider the problem of model building for rare events prediction in longitudinal follow-up studies. In this paper, we compare several resampling methods to improve standard regression models on a real life example. We evaluate the effect of the sampling rate on the predictive performances of the models. To evaluate the predictive performance of a longitudinal model, we consider a validation te…
▽ More
We consider the problem of model building for rare events prediction in longitudinal follow-up studies. In this paper, we compare several resampling methods to improve standard regression models on a real life example. We evaluate the effect of the sampling rate on the predictive performances of the models. To evaluate the predictive performance of a longitudinal model, we consider a validation technique that takes into account time and corresponds to the actual use in real life.
△ Less
Submitted 19 June, 2023;
originally announced June 2023.
-
Improper vs finitely additive distributions as limits of countably additive probabilities
Authors:
Erwan Saint Loubert Bié,
Pierre Druilhet,
Erwan Saint,
Loubert Bié
Abstract:
In Bayesian statistics, improper distributions and finitely additive probabilities (FAPs) are the two main alternatives to proper distributions, i.e. countably additive probabilities. Both of them can be seen as limits of proper distribution sequences w.r.t. to some specific convergence modes. Therefore, some authors attempt to link these two notions by this means, partly using heuristic arguments…
▽ More
In Bayesian statistics, improper distributions and finitely additive probabilities (FAPs) are the two main alternatives to proper distributions, i.e. countably additive probabilities. Both of them can be seen as limits of proper distribution sequences w.r.t. to some specific convergence modes. Therefore, some authors attempt to link these two notions by this means, partly using heuristic arguments. The aim of the paper is to compare these two kinds of limits. We show that improper distributions and FAPs represent two distinct characteristics of a sequence of proper distributions and therefore, surprisingly, cannot be connected by the mean of proper distribution sequences. More specifically, for a sequence of proper distribution which converge to both an improper distribution and a set of FAPs, we show that another sequence of proper distributions can be constructed having the same FAP limits and converging to any given improper distribution. This result can be mainly explained by the fact that improper distributions describe the behavior of the sequence inside the domain after rescaling, whereas FAP limits describe how the mass concentrates on the boundary of the domain. We illustrate our results with several examples and we show the difficulty to define properly a uniform FAP distribution on the natural numbers as an equivalent of the improper flat prior. MSC 2010 subject classifications: Primary 62F15; secondary 62E17,60B10.
△ Less
Submitted 19 June, 2023; v1 submitted 18 June, 2019;
originally announced June 2019.
-
On the Flatland Paradox
Authors:
Pierre Druilhet
Abstract:
We revisit the flatland paradox proposed by \cite{ston1976} which is an example of non-conglomerability. The aim of the paper is to show that the improperness of the prior is not directly involved in the inconsistency. First, we show that the choice of a flat prior is not adapted to the structure of the parameter space and we consider an improper prior based on reference priors with nuisance param…
▽ More
We revisit the flatland paradox proposed by \cite{ston1976} which is an example of non-conglomerability. The aim of the paper is to show that the improperness of the prior is not directly involved in the inconsistency. First, we show that the choice of a flat prior is not adapted to the structure of the parameter space and we consider an improper prior based on reference priors with nuisance parameter for which the Bayesian analysis matches the intuitive reasoning. Then, we propose an analysis by considering the flat prior as limit of proper uniform priors. In order to use limiting arguments, we must make a distinction between two different Bayesian paradigms. The first one is related to the marginal model whereas the second one is related to the conditional model. For the latter approach, we show that the inconsistency remains even with proper priors provided that we reconsider the interpretation of prior distributions.
△ Less
Submitted 7 April, 2017; v1 submitted 1 May, 2015;
originally announced May 2015.
-
Optimal cross-over designs for full interaction models
Authors:
R. A. Bailey,
P. Druilhet
Abstract:
We consider repeated measurement designs when a residual or carry-over effect may be present in at most one later period. Since assuming an additive model may be unrealistic for some applications and leads to biased estimation of treatment effects, we consider a model with interactions between carry-over and direct treatment effects. When the aim of the experiment is to study the effects of a trea…
▽ More
We consider repeated measurement designs when a residual or carry-over effect may be present in at most one later period. Since assuming an additive model may be unrealistic for some applications and leads to biased estimation of treatment effects, we consider a model with interactions between carry-over and direct treatment effects. When the aim of the experiment is to study the effects of a treatment used alone, we obtain universally optimal approximate designs. We also propose some efficient designs with a reduced number of subjects.
△ Less
Submitted 19 November, 2014; v1 submitted 4 November, 2014;
originally announced November 2014.
-
Approximation of improper priors
Authors:
Christele Bioche,
Pierre Druilhet
Abstract:
We propose a convergence mode for positive Radon measures which allows a sequence of probability measures to have an improper limiting measure. We define a sequence of vague priors as a sequence of probability measures that converges to an improper prior. We consider some cases where vague priors have necessarily large variances and other cases where they have not. We study the consequences of the…
▽ More
We propose a convergence mode for positive Radon measures which allows a sequence of probability measures to have an improper limiting measure. We define a sequence of vague priors as a sequence of probability measures that converges to an improper prior. We consider some cases where vague priors have necessarily large variances and other cases where they have not. We study the consequences of the convergence of prior distributions on the posterior analysis. Then we give some constructions of vague priors that approximate the Haar measures or the Jeffreys priors. We also revisit the Jeffreys-Lindley paradox.
△ Less
Submitted 30 March, 2016; v1 submitted 5 November, 2013;
originally announced November 2013.
-
Optimality of neighbor-balanced designs for total effects
Authors:
R. A. Bailey,
P. Druilhet
Abstract:
The purpose of this paper is to study optimality of circular neighbor-balanced block designs when neighbor effects are present in the model. In the literature many optimality results are established for direct effects and neighbor effects separately, but few for total effects, that is, the sum of direct effect of treatment and relevant neighbor effects. We show that circular neighbor-balanced de…
▽ More
The purpose of this paper is to study optimality of circular neighbor-balanced block designs when neighbor effects are present in the model. In the literature many optimality results are established for direct effects and neighbor effects separately, but few for total effects, that is, the sum of direct effect of treatment and relevant neighbor effects. We show that circular neighbor-balanced designs are universally optimal for total effects among designs with no self neighbor. Then we give efficiency factors of these designs, and show some situations where a design with self neighbors is preferable to a neighbor-balanced design.
△ Less
Submitted 5 October, 2004;
originally announced October 2004.
