|
|
|
 |
Search published articles |
 |
|
Showing 41 results for Regression
Mojtaba Khazaei,
Volume 2, Issue 1 (8-2008)
Abstract
One of the models that can be used to study the relationship between Boolean random sets and explanatory variables is growth regression model which is defined by generalization of Boolean model and permitting its grains distribution to be dependent on the values of explanatory variables. This model can be used in the study of behavior of Boolean random sets when their coverage regions variation is associated with the variation of grains size. In this paper we make possible the identification and fitting suitable growth model using available information in Boolean model realizations and values of explanatory variables. Also, a suitable method for fitting growth regression model is presented and properties of its obtained estimators are studied by a simulation study.
Maryam Torkzadeh, Soroush Alimoradi,
Volume 3, Issue 1 (9-2009)
Abstract
One of the tools for determining nonlinear effects and interactions between the explanatory variables in a logistic regression model is using of evolutionary product unit neural networks. To estimate the model parameters constructed by this method, a combination of evolutionary algorithms and classical optimization tools is used. In this paper, we change the structure of neural networks in the form that all model parameters can be estimated by using an evolutionary algorithms causes a model that is Akaike information criterion is better than conventional logisti model Akaike information criterion, but using the combination method gives the best model.
Mehdi Akbarzadeh, Hamid Alavimajd, Yadollah Mehrabi, Maryam Daneshpoor, Anvar Mohammadi,
Volume 3, Issue 2 (3-2010)
Abstract
One of the important problems that bring up in genetic fields is determining of loci of special gene in order to gene mapping and generating more effective drugs in medicine. Genetic linkage analysis is one important stage in this way. Haseman-Elston method is a quantitative statistical method that is used by biostatisticians and geneticists for genetic linkage analysis. The original Haseman-Elston method is presented in the year 1972 and ever after many investigators recommended some suggestions to make better it. In this article, we introduce the Haseman-Elston regression method and its extensions through 1972 to 2009. and finally we show performance of these methods in a practical example.
Mehrdad Niaparast, Sahar Mehr-Mansour,
Volume 4, Issue 1 (9-2010)
Abstract
The main part of optimal designs in the mixed effects models concentrates on linear models and binary models. Recently, Poisson models with random effects have been considered by some researchers. In this paper, an especial case of the mixed effects Poisson model, namely Poisson regression with random intercept is considered. Experimental design variations are obtained in terms of the random effect variance and indicated that the variations depend on the variance parameter. Using D-efficiency criterion, the impression of random effect on the experimental setting points is studied. These points are compared with the optimal experimental setting points in the corresponding model without random effect. We indicate that the D-efficiency depends on the variance of random effect.
Ebrahim Khodaie, Roohollah Shojaei,
Volume 6, Issue 1 (8-2012)
Abstract
Sampling weights are calibrated according to the theory of calibration when the sum of population total for auxiliary variables is known. Under known population, totals for auxiliary variables and some conditions Devile and Sarndal showed that generalized regression estimators could approximate calibration estimators and their variances. In this paper, under unknown population totals for auxiliary variables, an estimator for the population total is proposed and its variance is obtained. It is shown that our estimator for the population total is more efficient than the Horvitz-Thompson estimators by theoretically and simulation results.
Arezou Mojiri, Soroush Alimoradi, Mohammadreza Ahmadzade,
Volume 7, Issue 1 (9-2013)
Abstract
Logistic regression models in classification problems by assuming the linear effects of covariates is a modeling for class membership posterior probabilities. The main problem that includes nonlinear combinations of covariates is maximum likelihood estimation (MLE) of the model parameters. In recent investigations, an approach of solving this problem is combination of neural networks, evolutionary algorithms and MLE methods. In this paper, another type of radial basis functions, namely inverse multiquadratic functions and hybrid method, are considered for estimating the parameters of these models. The experimental results of comparing the proposed models show that the inverse multiquadratic functions compared to the Gaussian functions have better precision in classification problems.
Kobra Gholizadeh, Mohsen Mohammadzadeh, Zahra Ghayyomi,
Volume 7, Issue 1 (9-2013)
Abstract
In Bayesian analysis of structured additive regression models which are a flexible class of statistical models, the posterior distributions are not available in a closed form, so Markov chain Monte Carlo algorithm due to complexity and large number of hyperparameters takes long time. Integrated nested Laplace approximation method can avoid the hard simulations using the Gaussian and Laplace approximations. In this paper, consideration of spatial correlation of the data in structured additive regression model and its estimation by the integrated nested Laplace approximation are studied. Then a crime data set in Tehran city are modeled and evaluated. Next, a simulation study is performed to compare the computational time and precision of the models provided by the integrated nested Laplace approximation and Markov chain Monte Carlo algorithm
Afshin Fallah, Mahsa Nadifar, Ramin Kazemi,
Volume 7, Issue 1 (9-2013)
Abstract
In this paper the regression analysis with finite mixture bivariate poisson response variable is investigated from the Bayesian point of view. It is shown that the posterior distribution can not be written in a closed form due to the complexity of the likelihood function of bivariate Poisson distribution. Hence, the full conditional posterior distributions of the parameters are computed and the Gibbs algorithm is used to sampling from posterior distributions. A simulation study is performed in order to assess the proposed Bayesian model and its efficiency in estimation of the parameters is compared with their frequentist counterparts. Also, a real example presented to illustrate and assess the proposed Bayesian model. The results indicate to the more efficiency of the estimators resulted from Bayesian approach than estimators of frequentist approach at least for small sample sizes.
Sedighe Zamani Mehryan, Ali Reza Nematollahi,
Volume 7, Issue 2 (3-2014)
Abstract
In this paper, the pseudo-likelihood estimators and the limiting distribution of the score test statistic associated with several hypothesis tests such as unit root test for the linear regression models with stationary and nonstationary residuals are calculated. The limiting behavior of theses test statistics by using a simpler approach of the original presentation is derived. Also by using a Mont Carlo method, it is shown that the derived pseudo-likelihood estimators are appropriate. The quantiles of the limiting distribution of the test statistic for a unit root are also calculated and a new table is provided which can be used by researchers for the unit root test.
Jalal Chachi, Gholamreza Hesamian,
Volume 8, Issue 1 (9-2014)
Abstract
In this paper, we deal with modeling crisp input-fuzzy output data by constructing a MARS-fuzzy regression model with crisp parameters estimation and fuzzy error terms for the fuzzy data set. The proposed method is a two-phase procedure which applies the MARS technique at phase one and an optimization problem at phase two to estimate the center and fuzziness of the response variable. A realistic application of the proposed method is also presented in a hydrology engineering problem. Empirical results demonstrate that the proposed approach is more efficient and more realistic than some traditional least-squares fuzzy regression models.
Sahar Mehrmansour, Mehrdad Niaparast,
Volume 8, Issue 2 (3-2015)
Abstract
The main researches of optimum experimental designs for mixed effects have been concentrated on locally optimal designs. These designs are obtained based on the initial guess of parameters. Therefore, locally designs may be the best design but for wrong assumed model. Recently, Bayesian approach has been considered by researches when information about model parameters is available. In the present work, optimal design for the mixed effects Poisson regression model based on some prior distributions are considered and for two special cases of this models the Bayesian D-optimal designs are obtained for some representative values of variance of random effect. The results are compared to Poisson regression model without random effects.
Ehsan Zamanzade,
Volume 8, Issue 2 (3-2015)
Abstract
In this paper, an improved mean estimator for unbalanced ranked set samples is proposed. The estimator is obtained by using the fact that distribution function of order statistics are stochastically ordered. Also, it is showed that this estimator is convergent and has better performance than its empirical counterpart in unbalanced ranked set samples.
Roshanak Aliakbari Saba, Alireza Zahedian, Marzieh Arbabi,
Volume 9, Issue 1 (9-2015)
Abstract
Annual estimation of average household incomes is one of the main goals of the household income and expenditure survey in Iran. So, regarding importance of accuracy of gathered data and reasons that lead to error in measuring household income, in this paper, model-based methods are used for estimating income measurement error and adjusting sample households declared income for 2011 household income and expenditure survey.
Ali Aghamohammadi, Sakineh Mohammadi,
Volume 9, Issue 2 (2-2016)
Abstract
In many medical studies, in order to describe the course of illness and treatment effects, longitudinal studies are used. In longitudinal studies, responses are measured frequently over time, but sometimes these responses are discrete and with two-state. Recently Binary quantile regression methods to analyze this kind of data have been taken into consideration. In this paper, quantile regression model with Lasso and adaptive Lasso penalty for longitudinal data with dichotomous responses is provided. Since in both methods posteriori distributions of the parameters are not in explicit form, thus the full conditional posteriori distributions of parameters are calculated and the Gibbs sampling algorithm is used to deduction. To compare the performance of the proposed methods with the conventional methods, a simulation study was conducted and at the end, applications to a real data set are illustrated.
Abouzar Bazyari,
Volume 10, Issue 1 (8-2016)
Abstract
Hypothesis testing the homogeneity of means of k univariate normal populations against the hypothesis of one sided ordered means with unknown and equal variances is considered. A new completely method to find the uniformly most powerful test at significance level α is presented based on the multivariate t distribution. Since for more than two populations finding the null distribution of test statistic is not easy, the power of test is computed and then the critical values of test statistic for different significance levels obtained. This testing method is used for real examples. Also testing homogeneity of k mean vectors against two sided ordered mean vectors of multivariate normal populations is considered. Using Monte Carlo simulation the values of classical power of test for two bivariate and trivariate normal distributions at different significance levels are compared.
Habib Jafari, Shima Pirmohamadi,
Volume 10, Issue 2 (2-2017)
Abstract
The optimal criteria are used to find the optimal design in the studied model. These kinds of models are included the paired comparison models. In these models, the optimal criteria (D-optimality) determine the optimal paired comparison. In this paper, in addition to introducing the quadratic regression model with random effects, the paired comparison models were presented and the optimal design has been calculated for them.
Mahtab Tarhani, Sayed Mohammad Reaz Alavi,
Volume 10, Issue 2 (2-2017)
Abstract
In weighted sampling as a generalization of random sampling, every observation, y, is recorded with probably proportional to a non-negative function of y. In this paper, the normal regression model is investigated under the weighted sampling for a common weight function. Parameters of the model are estimated for known and unknown weight parameters. Using simulation, efficiency of estimators is studied when they have not closed forms. As an application, the data of number of visited patients by specialist doctors in Social Security Organization of Ahvaz in Iran (SSOAI) are analyzed.
Ali Aghamohammadi, Mahdi Sojoudi,
Volume 10, Issue 2 (2-2017)
Abstract
Value-at-Risk and Average Value-at-Risk are tow important risk measures based on statistical methoeds that used to measure the market's risk with quantity structure. Recently, linear regression models such as least squares and quantile methods are introduced to estimate these risk measures. In this paper, these two risk measures are estimated by using omposite quantile regression. To evaluate the performance of the proposed model with the other models, a simulation study was conducted and at the end, applications to real data set from Iran's stock market are illustarted.
Omid Akhgari, Mousa Golalizadeh,
Volume 10, Issue 2 (2-2017)
Abstract
The presence of endogenous variables in the statistical models leads to inconsistent and bias estimators for the parameters. In this case, several approaches have been proposed which are able to tackle the biase and inconsistency problems only in large sample situations. One of these methods is biased on instrumental variables which causes removing endogenous variables. The method of two-stage least squares is another approach in this case that it has more accurate than ordinary least squares. This paper aims to enhance the accuracy of three methods of estimation based upon least square methodology called, two-stage iterative least squares, two-stage Jackknife least squares and also two-stage calibration least squares. In order to evaluate the performance of each method, a simulation study is conducted. Also, using data collected in 1390 related to the cost and revenue in Iran, those methods to estimate parameters are compared.
Maliheh Heidari, Farzad Eskandari,
Volume 11, Issue 1 (9-2017)
Abstract
In this paper the issue of variable selection with new approach in finite mixture of semi-parametric regression models is studying, although it is supposed that data have Poisson distribution. When we use Poisson distribution, two problems such as overdispersion and excess zeros will happen that can affect on variable selection and parameter estimation. Actually parameter estimation in parametric component of the semi-parametric regression model is done by penalized likelihood approach. However, in nonparametric component after local approximation using Teylor series, the estimation of nonparametric coefficients along with estimated parametric coefficients will be calculated. Using new approach leads to a properly variable selection results. In addition to representing related theories, overdispersion and excess zeros are considered in data simulation section and using EM algorithm in parameter estimation leads to increase the accuracy of end results.
|
|
