Australian Journal of Basic and Applied Sciences
        
   

Jan-Feb 2024

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Evaluation and Comparison of Outliers’ Detection Methods: A Simulation Study


Jhennifer dos Santos Nascimento, Jaqueline Akemi Suzuki Sediyama, Anderson Cristiano Neisse, Thaynara Aparecida de Souza Neto, Paulo César Emiliano, José Ivo Ribeiro Júnior, Paulo Roberto Cecon

ABSTRACT: Detecting outliers is crucial for a deeper understanding of the data set. It helps select robust methods and should precede any type of statistical data analysis. In this context, the present work aimed to evaluate and compare three outlier and four leverage point detection methods using simulation. For leverage point detection, the Leverage, DFFIT, DFBETA, and Cook distance methods were considered. For outlier detection, three variations of the Mahalanobis distance were used: the classical, the minimum covariance determinant (MCD), and the minimum volume ellipsoid (MVE). The simulation study was performed in R software, in which the methods were compared by evaluating the proportion of total detection and proportion of correct detection of different groups of discrepant points inserted into a database containing ten predictor variables and one response variable, with 100 observations each. Finally, to corroborate the results obtained in the simulation study, the seven methods were used to detect discrepant points in a real database from the milk production chain. The studies showed that the classical Mahalanobis distance and the Leverage method presented the closest evaluated detection proportions to one and were considered the best detection methods.

[ FULL TEXT PDF 1-12 ] DOI: 10.22587/ajbas.2024.18.1.1

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Morpho-Pedological Diagnosis Of The Hills In South-West Of Côte d’Ivoire: Case Of An Open Trench at University of San Pedro

Ouattara Amidou, Kone Brahima, Konan Kouassi Urbain

ABSTRACT: The very rugged relief of San Pedro region restricts the rational use of land for the agricultural activities on which people depend. This situation will result in a long-term threat to food and economic security of the region's populations. The aim of study was therefore, to suggest to farmers a primary approach to efficient use of ferralsol of San Pedro hills for their activities. To achieve this, a morpho-pedological study of a trench opened up during development work for building University of San Pedro was carried out, based on relevant soil description criteria, as used by Overseas Scientific and Technical Research Office. The study highlighted four major horizons in the weathering profile, and the major agronomic constraints to their development, including induration and a high load of coarse elements. In addition, these soils have a clay texture that makes them threshing in the rainy season and pulverising in the dry season. To make the best use of these soils, amendments and development of the indurated horizons are recommended. The adoption of contour cultivation is proposed to limit soil erosion on hillsides. On other hand, on higher ground (slopes greater than 30°C), the installation of grazing land was recommended. The students involved in this study were able to consolidate the knowledge they had acquired in the general pedology lecture course.

[ FULL TEXT PDF 13-18 ] DOI: 10.22587/ajbas.2024.18.1.2

--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

   
   
   
    ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------