Afrika Statistika https://www.ajol.info/index.php/afst The Africa Statistika journal publishes applied and theoretical work researches about probabilities, statistics, operational research, financial mathematics and about the annexes as well. The fundamentals articles will be published in a series A, the applied articles in a series B. Each series will be displayed at least once per year. The articles are written in English or in French and have to be typed out with double spacing on the first side only, with a margin of three centimetres. The articles must not be longer than 12 pages. The invited papers can be longer. Other websites related to this journal:<a title=" www.jafristat.net" href="http://univi.net/jas/" target="_blank"> www.jafristat.net</a> en-US Copyright is owned by the journal ganesamblo@yahoo.com (Prof. Gane Samb Lo) mdoucamara@gmail.com (Mamadou Camara) Thu, 21 Nov 2024 10:16:24 +0000 OJS 3.3.0.11 http://blogs.law.harvard.edu/tech/rss 60 On some inferences of Lévy distribution https://www.ajol.info/index.php/afst/article/view/283054 <p>The L<span style="font-weight: 400;">é</span>vy distribution is one of the three stable distributions that has ´ probability density function in simple closed form. This distribution is used in modeling stock prices. In this paper, we present some properties of this distribution. Based on the basic properties some characterizations of this distribution are given.</p> <p>La loi de probabilite L<span style="font-weight: 400;">é</span>vy figurent parmi les lois stables ayant un expression explicite de la densit<span style="font-weight: 400;">é</span> de probabilit<span style="font-weight: 400;">é </span>. Elle est souvent utilis<span style="font-weight: 400;">é</span>e pour mod<span style="font-weight: 400;">é</span>liser le prix des actions en Finance. Dans ce papier, nous pr<span style="font-weight: 400;">é</span>sentos quelques de ses proprietes <span style="font-weight: 400;">á</span> partir desquelles des charact<span style="font-weight: 400;">é</span>risations sont donn<span style="font-weight: 400;">é</span>es.</p> Mohammad Ahsanullah, Valery B. Nevzorov Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283054 Thu, 21 Nov 2024 00:00:00 +0000 A dynamic markov regime-switching asymmetric GARCH model and its cumulative impulse response function https://www.ajol.info/index.php/afst/article/view/283055 <p>In this paper, we consider the Markov regime-switching GJRGARCH(1,1) model to capture both the cumulative impulse response and the asymmetry of the dynamic behavior of financial market volatility in stationary and explosive states. The model can capture regime shifts in volatility between two regimes as well as the asymmetric response to negative and positive shocks. A Monte Carlo simulation is conducted to validate the main theory and find that the regime-switching GJR-GARCH model performs better than the standard GJRGARCH model. Applications to Brazilian stock market data show that the proposed model performs well in terms of cumulative impulse response.</p> <p>Dans cet article, nous examinons le mod<span style="font-weight: 400;">è</span>le GJR-GARCH(1,1) <span style="font-weight: 400;"> à</span> changement de r<span style="font-weight: 400;">é</span>gime de Markov pour capturer <span style="font-weight: 400;">à</span> la fois la r<span style="font-weight: 400;">é</span>ponse impulsionnelle cumulative et l’asym<span style="font-weight: 400;">é</span>trie du comportement dynamique de la volatilit<span style="font-weight: 400;">é</span> des march<span style="font-weight: 400;">é</span>s financiers dans les <span style="font-weight: 400;">é</span>tats stationnaires et explosifs. Le mod<span style="font-weight: 400;">è</span>le peut capturer les changements de r<span style="font-weight: 400;">é</span>gime de la volatilit<span style="font-weight: 400;">é</span> entre deux r<span style="font-weight: 400;">é</span>gimes ainsi que la r<span style="font-weight: 400;">é</span>ponse asym<span style="font-weight: 400;">é</span>trique aux chocs n<span style="font-weight: 400;">é</span>gatifs et positifs. Une simulation de Monte Carlo est men<span style="font-weight: 400;">ée</span> pour valider la th<span style="font-weight: 400;">é</span>orie principale et trouver que le mod<span style="font-weight: 400;">é</span>le GJR-GARCH <span style="font-weight: 400;">á</span> changement de r<span style="font-weight: 400;">é</span>gime est plus performant que le mod<span style="font-weight: 400;">è</span>le GJR-GARCH standard. Les applications aux donn<span style="font-weight: 400;">é</span>es du march<span style="font-weight: 400;">é</span> boursier br<span style="font-weight: 400;">é</span>silien montrent que le mod<span style="font-weight: 400;">è</span>le propos<span style="font-weight: 400;">é</span> est performant en termes de r<span style="font-weight: 400;">é</span>ponse impulsionnelle cumulative.</p> Gado Sema, Mamadou Abdoulaye Konté, Abdou Kâ Diongue Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283055 Thu, 21 Nov 2024 00:00:00 +0000 Stein’s method in two limit theorems involving the generalized inverse Gaussian distribution https://www.ajol.info/index.php/afst/article/view/283056 <p>On one hand, the generalized hyperbolic (<em>GH</em>) distribution converges in law to the generalized inverse Gaussian (<em>GIG</em>) distribution under certain conditions on the parameters. On the other hand, when the edges of an infinite rooted tree are equipped with independent resistances whose distributions are inverse Gaussian or reciprocal inverse Gaussian distributions, the total resistance converges almost surely to some random variable which follows the reciprocal inverse Gaussian (<em>RIG</em>) distribution. In this paper we provide explicit upper bounds for the distributional distance between the <em>GH</em> distribution (resp. the distribution of the total resistance of the tree) and their limiting GIG (resp. <em>RIG</em>) distribution applying Stein’s method.</p> <p>Sous certaines conditions sur ses param<span style="font-weight: 400;">è</span>tres, la loi hyperbolique g<span style="font-weight: 400;">é</span>n<span style="font-weight: 400;">é</span>ralis<span style="font-weight: 400;">é</span> (<em>GH</em>) converge vers la loi gaussienne inverse g<span style="font-weight: 400;">é</span>n<span style="font-weight: 400;">é</span>ralis<span style="font-weight: 400;">é</span>e (<em>GIG</em>). Lorsque les ar<span style="font-weight: 400;">ê</span>tes d’un arbre infini sont munies de r<span style="font-weight: 400;">é</span>sistances al<span style="font-weight: 400;">é</span>atoires ind<span style="font-weight: 400;">é</span>pendantes, de loi gaussienne inverse ou de loi gaussienne inverse r<span style="font-weight: 400;">é</span>ciproque, la r<span style="font-weight: 400;">é</span>sistance <span style="font-weight: 400;">é</span>quivalente converge presque s<span style="font-weight: 400;">û</span>rement vers une variable al<span style="font-weight: 400;">é</span>atoire de loi gaussienne inverse r<span style="font-weight: 400;">é</span>ciproque (<em>RIG</em>). Dans cet article, nous d<span style="font-weight: 400;">é</span>terminons des majorants explicites de la distance probabiliste entre la loi <em>GH</em> (resp. la loi d’un circuit arborescent) et la loi limite <em>GIG</em> (resp. <em>RIG</em>) en appliquant la m<span style="font-weight: 400;">é</span>thode de Stein.</p> Essomanda Konzou, Efoévi Koudou , Kossi Essona Gneyou Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283056 Thu, 21 Nov 2024 00:00:00 +0000 Log-exponential power distribution for accelerated failure time model in survival analysis and its application https://www.ajol.info/index.php/afst/article/view/283057 <p>We proposed the log-exponential power density function as baseline distribution for accelerated failure time model (<em>AFT</em>) used in analysis of survival data with covariates. This model generalizes the log-normal and some exponential family due to flexibility at the tail region. It has log-concavity property, accommodates the four basic shapes of hazard function which is an attractive property compared with other distributions that cannot accommodate same. The model’s goodness of fit relative to some existing models was tested using data from chronic liver disease patients monitored at Obafemi Awolowo University Teaching Hospital, Ile-Ife.</p> <p>Dans ce papier, nous proposons la fonction de densit<span style="font-weight: 400;">é</span> de puissance log-exponentielle comme distribution comme fonction de base pour le mod<span style="font-weight: 400;">è</span>le du temps d<span style="font-weight: 400;">é</span>faillance acc<span style="font-weight: 400;">é</span>l<span style="font-weight: 400;">é</span>r<span style="font-weight: 400;">é </span><em>AFT</em> en Analyse des donn<span style="font-weight: 400;">é</span>es de survie avec des covariables. Ce modèle g<span style="font-weight: 400;">é</span>n<span style="font-weight: 400;">é</span>ralise la famille log-normale et une famille exponentielle en raison de la flexibilit<span style="font-weight: 400;">é</span> de la queue de la distribution. Ce mod<span style="font-weight: 400;">è</span>le a une propri<span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span> de concavit<span style="font-weight: 400;">é</span> logarithmique, s’adapte aux quatre formes de base de la fonction de danger, ce qui est une propri<span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span> attrayante par rapport aux autres distributions qui n’ont pas. La qualit<span style="font-weight: 400;">é</span> de l’ajustement du mod<span style="font-weight: 400;">è</span><span style="font-weight: 400;">le</span> par rapport <span style="font-weight: 400;">à</span> certains mod<span style="font-weight: 400;">è</span>les existants a <span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span> test<span style="font-weight: 400;">é</span>e <span style="font-weight: 400;">à</span> l’aide de donn<span style="font-weight: 400;">é</span>es provenant de patients atteints d’une maladie h<span style="font-weight: 400;">é</span>patique chronique suivis <span style="font-weight: 400;">à</span> l’h<span style="font-weight: 400;">ô</span>pital universitaire Obafemi Awolowo, Ile-Ife.</p> Akinlolu Adeseye Olosunde, Chidimma Florence Ejiofor Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283057 Thu, 21 Nov 2024 00:00:00 +0000 The Marshall-Olkin-Gumbel extended Weibull distribution: Properties and applications https://www.ajol.info/index.php/afst/article/view/283058 <p>We introduce a new lifetime distribution called Marshall-Olkin extended Gumbel-Weibull. Some properties of distribution such as moments, TLmoments, quantile function, entropy, and order statistics are studied. The fexibility of the distribution to model unimodal, monotone shapes as well as unimodal, bimodal, monotone failure rates are presented. The estimators of the parameters of the distribution were obtained using the maximum likelihood estimation method. The performance of the maximum likelihood estimates of the Marshall-Olkin extended Gumbel-Weibulll parameters was observed through simulation studies.Two real life applications to illustrate the potentials of the new distribution are presented, and comparison with other distribution having the same baseline is done using goodness-of-test statistics.</p> <p>Nous introduisons une nouvelle distribution pour mod<span style="font-weight: 400;">é</span>liser des dur<span style="font-weight: 400;">é</span>e de vie appel<span style="font-weight: 400;">é</span>e extension Marshall-Olkin de la loi de Gumbel-Weibull. Certaines propri<span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span>s de la distribution telles que les moments, les moments TL, la fonction quantile, l’entropie et les statistiques d’ordre sont <span style="font-weight: 400;">é</span>tudi<span style="font-weight: 400;">é</span>es. La fexibilit<span style="font-weight: 400;">é</span> de la distribution pour mod<span style="font-weight: 400;">é</span>liser des formes unimodales et monotones ainsi les taux de d<span style="font-weight: 400;">é</span>faillance unimodaux, bimodaux et monotones sont pr<span style="font-weight: 400;">é</span>sent<span style="font-weight: 400;">é</span>s. Les estimateurs des param<span style="font-weight: 400;">è</span>tres de la distribution ont <span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span> obtenus <span style="font-weight: 400;">à</span> l’aide de la m<span style="font-weight: 400;">é</span>thode d’estimation du maximum de vraisemblance. La performance des estimations du maximum de vraisemblance des param<span style="font-weight: 400;">è</span>tres de Gumbel-Weibulll <span style="font-weight: 400;">é</span>tendus de Marshall-Olkin a <span style="font-weight: 400;">é</span>t<span style="font-weight: 400;">é</span> observ<span style="font-weight: 400;">é</span>e au moyen d’<span style="font-weight: 400;">é</span>tudes de simulation. Deux applications r<span style="font-weight: 400;">é</span>elles pour illustrer les potentiels de la nouvelle distribution sont pr<span style="font-weight: 400;">é</span>sent<span style="font-weight: 400;">é</span>es, et la comparaison avec d’autres distributions ayant la m<span style="font-weight: 400;">ê</span>me base de r<span style="font-weight: 400;">é</span>f<span style="font-weight: 400;">é</span>rence est effectu<span style="font-weight: 400;">é</span>e <span style="font-weight: 400;">à</span> l’aide de tests d’ajustage.</p> Elebe Emmanuel Nwezza, Fidelis Ifeanyi Ugwuowo Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283058 Thu, 21 Nov 2024 00:00:00 +0000 Adaptive Realized Hyperbolic GARCH process: Stability and estimation https://www.ajol.info/index.php/afst/article/view/283062 <p>In this paper, we propose an Adaptive Realized Hyperbolic <em>GARCH</em> (ARealized <em>HYGARCH</em>) process to model the long memory of high-frequency time series with possible structural breaks. The structural change is modeled by allowing the intercept to follow the smooth and flexible function form introduced by Gallant (1984). In addition, stability conditions of the process are investigated. A Monte Carlo study is considered in order to illustrate the performance of the ARealized <em>HYGARCH</em> process compared to the Realized <em>HYGARCH</em> with or without structural change.</p> <p>Dans cet article, nous proposons un mod<span style="font-weight: 400;">è</span>le hyperbolique GARCH r<span style="font-weight: 400;">é</span>alis<span style="font-weight: 400;">é </span>adaptatif (A-Realized HYGARCH) pour mod<span style="font-weight: 400;">é</span>liser la longue m<span style="font-weight: 400;">é</span>moire des s<span style="font-weight: 400;">é</span>ries chronologiques <span style="font-weight: 400;">á</span> haute fr<span style="font-weight: 400;">é</span>quence avec d’<span style="font-weight: 400;">é</span>ventuelles changements de r<span style="font-weight: 400;">é</span>gimes. Le changement de r<span style="font-weight: 400;">é</span>gime est mod<span style="font-weight: 400;">é</span>lis<span style="font-weight: 400;">é</span>, en permettant l’intercepte de suivre une forme de fonction lisse et flexible introduite par Gallant (1984). De plus, les conditions de stabilit<span style="font-weight: 400;">é</span> pour ce mod<span style="font-weight: 400;">è</span>le sont <span style="font-weight: 400;">é</span>tablies dans ce papier. Une <span style="font-weight: 400;">é</span>tude de Monte Carlo est consid<span style="font-weight: 400;">é</span>r<span style="font-weight: 400;">é</span>e afin d’illustrer les performances du mod<span style="font-weight: 400;">è</span>le (A-Realized <em>HYGARCH</em>) compar<span style="font-weight: 400;">é</span> au mod<span style="font-weight: 400;">è</span>le <em>HYGARCH</em> Realis<span style="font-weight: 400;">é</span> sur des donn<span style="font-weight: 400;">é</span>es avec ou sans changement structurel.</p> El Hadji Mamadou Sall, El Hadji Deme, Abdou Kâ Diongue Copyright (c) 2024 https://www.ajol.info/index.php/afst/article/view/283062 Thu, 21 Nov 2024 00:00:00 +0000