نشریه گستره علوم آماری
سال یکم شماره 2 (بهار و تابستان 1395)

This paper describes the performance analysis of two classifier models common in statistics and data mining on binary dependent variable, binary Logistic Regression (B.LR) and Classification Regression Tree (CART). The evaluation method is using all data in training stage. The using data set is from “Evaluation of patients with Jaundice on children” report. Data set is collection of categorical and continues independent variables. The classification performance of two classifiers is presented by using statistical performance measures like accuracy, specificity and sensitivity. Experimental result showed that accuracy of LR is more than 83% and CLASSIFICATION AND REGRESSION TREE is nearly 73%. So the sensitivity measure for BINARY LOGISTIC REGRESSION is nearby 77% and 66% for CLASSIFICATION AND REGRESSION TREE as well the specificity scale is 85% for BINARY LOGISTIC REGRESSION and 76% for CLASSIFICATION AND REGRESSION TREE. The result shows the performance of BINARY LOGISTIC REGRESSION classifier is found to be better than CLASSIFICATION AND REGRESSION TREE.

In some statistical problems we need independence of two statistic, especially in testin hypotheses. With Basu's theorem, we can proof independence of two statistic without calculate their joint ditributions. In this paper we will give applications of this theorem in statistical testin hypothese as GLRT, UMPUT, independens LRT in Compound Tests, and testing of Econometric Models.

Wavelet Analysis is a branch of Harmonic Analysis and a new phenomenon of Mathematics science which offers wide range of application in Mathematics, Statistics and other fields. Wavelets analysis is finding a rapidly growing number of applications despite its young age and often replacing the conventional Fourier transform. Basically in this paper, the problem of estimating a density and its derivatives for a sample of censored random variables is considered. The purpose of this paper is to present an approach to this problem based on wavelets methods for extended negatively dependent observations. Besides, we explore its performances under the risk in Besov ball.

Line transect sampling is a very helpful method for estimating the density function of population in biology. In this paper, first the line transect sampling is introduced and then parametric hypothesis testing for half normal density function against univariate exponential density function in line transect sampling was considered.
The test statistics is obtained using likelihood ratio method. Computing the null distribution of test statistic and its critical values is not easy because of the complexity of test statistic structure, therefore Monte carlo simulation was used for finding the critical values of test statistic at different significance levels. This problem of testing was investigated with numerical examples.

This article first introduces the sterling polynomials and expresses the calculation methods in software R and their application in resolving a class of integrals which play an important role in the calculation of torques, torques of order statistics, Renyi and Shannon entropies in statistical distributions based on this type of integrals. Then, we obtain the torques of the two statistical distributions related to these integrals.

In this article, we introduce the Kumaraswamy Weibull distribution and discuss some features of this distribution.
That is a quite flexible model in analyzing positive data. It contains special sub-models the exponentiated Weibull, exponentiated Rayleigh,exponentiated exponential,Weibull and also the new Kumaraswamy exponential distribution, Explicit Expressions for moments, moment generating function, and also we present the results of the simulation on the distribution.

In this paper for the first time we introduce the Pseudo-moment generator function and also we investigate some properties of. .
We prove some properties of Pseudo-expectation and Pseudo-dispersion .

Correlation structure of spatial data is related to their positions and distances. Then spatial data analysis has various applications in applied areas.
In this research, we study two spatial interpolation methods, that is Ordinary Kriging and Universal Kriging. In this purpose, data sets are simulated and the performance of each of the methods is studied under the considered conditions. Besides, different sample sizes and variograms is considered to assess the effects of various sample sizes and different vaiogram functions. Furthermore, for every set of simulated data, cross validation criteria is applied to compare the methods. As an application of the results, the methods are applied for data of a mine.