References of "Fonton, N. H"
     in
Bookmark and Share    
See detailSur l'erreur résiduelle obtenue en régression linéaire multiple.
Akossou, A. Y. J.; Palm, Rodolphe ULg; Fonton, N. H. et al

Scientific conference (2007)

The residual error is one of the most used criteria for the choice of regression model. The large majority of the criteria for model selection are also functions of the usual variance estimate for a ... [more ▼]

The residual error is one of the most used criteria for the choice of regression model. The large majority of the criteria for model selection are also functions of the usual variance estimate for a regression model, as Akaike criterion information [Akaike, 1973] and Mallows’ criterion [Mallows, 1964]. The effectiveness of these criteria to approach the theoretical model is related to the good estimate of the theoretical residual error of the model. This efficiency is often limited by the sample size used for the estimates. The choice of a invalid model can have bad consequences on the objective of the research in particular on the forecasts, interpretations and the conclusions. We studied by Monte Carlo simulation the effects of the data structure on the quality of the residual error obtained in multiple linear regression. The results showed that the estimator is a good estimator when the model is established without selection of variables, but becomes biased when the model results from a selection of variables. Colinearity between the explanatory variables and the index which measure the importance of the variables in the true model do not have an influence on the quality of the estimator. [less ▲]

Detailed reference viewed: 122 (7 ULg)
Full Text
Peer Reviewed
See detailComparaison empirique des estimateurs des taux d'erreur en analyse discriminante
Glele Kakai, R.; Piraux, F.; Fonton, N. H. et al

in Revue de Statistique Appliquée (2003), LI(3), 61-74

Monte Carlo experiments are performed to compare twenty estimators or error rates in discriminant analysis in the case of two populations and linear classification rule. Several distributions (normal, chi ... [more ▼]

Monte Carlo experiments are performed to compare twenty estimators or error rates in discriminant analysis in the case of two populations and linear classification rule. Several distributions (normal, chi-square and beta) have been considered. Estimators eOS (for actual and expected error rate) and eB (for optimal error rate) are the best, except for distributions very different from normal distributions. The e632 estimator is the best non parametric estimator. This estimator is better than parmetric estimators for distributions very different from normal distributions. [less ▲]

Detailed reference viewed: 16 (0 ULg)
See detailIntroduction à la programmation avec Matlab sous windows
Akossou, A. Y. J.; Fonton, N. H.; Claustriaux, Jean-Jacques ULg

in Notes de Biométrie et d'Informatique (2001), 2

This note aims to introduce Matlab software. Through this presentation we show a few advantages of the software. This presentation is then illustrated by a few examples of program with the software.

Detailed reference viewed: 32 (2 ULg)
Full Text
See detailLa gestion des données dans le système SAS sous Windows v6.12
Fonton, N. H.; Claustriaux, Jean-Jacques ULg

in Notes de biométrie et d'informatique (2001), 1

The management of data is one of the major difficulties of the users in the SAS system. The note presents the procedures of data management after a brief summary on the organization and creation of data ... [more ▼]

The management of data is one of the major difficulties of the users in the SAS system. The note presents the procedures of data management after a brief summary on the organization and creation of data. Some procedures for statistics analysis are also presented. [less ▲]

Detailed reference viewed: 13 (2 ULg)
Full Text
Peer Reviewed
See detailModèle de productivité et tarif de cubage des peuplements de teck (tectona grandis L. F.) au Sud-Bénin
Fonton, N. H.; Palm, Rodolphe ULg; Akossou, A. Y. J. et al

in Annales des Sciences Agronomiques du Bénin (2001), 2(2), 209-224

This study aims to build site index model and stem volume function of pure, even aged and relatively homogeneous teak stands in south Benin. The characteristics of the stands were obtained from the ... [more ▼]

This study aims to build site index model and stem volume function of pure, even aged and relatively homogeneous teak stands in south Benin. The characteristics of the stands were obtained from the inventory of 104 temporary plots sampled from stands with different ages and fertility. The stem analysis of 19 dominant trees wer used to establish, by serial regression, a model giving the evolution of top height according to the age. The site index curves were built in order to determine site index of the stands. The estimated volume function obtained shows that age, site index and basal area are the main stand parameters of stem volume funtion. [less ▲]

Detailed reference viewed: 39 (1 ULg)
Full Text
Peer Reviewed
See detailComparaison empirique de méthodes de prédiction en régression linéaire multiple
Fonton, N. H.; Palm, Rodolphe ULg

in Revue de Statistique Appliquée (1998), XLVI(3), 53-64

Detailed reference viewed: 7 (0 ULg)