|
|
|
 |
Search published articles |
 |
|
General users only can access the published articles
Showing 10 results for Subject:
Ahmad Parsian, Shahram Azizi Sazi,
Volume 2, Issue 1 (8-2008)
Abstract
In this paper, a new class of estimators namely Constrained Bayes Estimators are obtained under Balanced Loss Function (BLF) and Weighted Balanced Loss Function (WBLF) using a ``Bayesian solution". The Constrained Bayes Estimators are calculated for the natural parameter of one-parameter exponential families of distributions. A common approach to the prior uncertainty in Bayesian analysis is to choose a class $Gamma$ of prior distributions and look for an optimal decision within the class $Gamma$. This is known as robust Bayesian methodology. Among several methods of choosing the optimal rules in the context of the robust Bayes method, we discuss obtaining Posterior Regret Constrained Gamma-Minimax (PRCGM) rule under Squared Error Loss and then employing the ``Bayesian solution", we obtain the optimal rules under BLF and WBLF.
Nasim Ejlali, Hamid Pezeshk,
Volume 2, Issue 2 (2-2009)
Abstract
Hidden Markov models are widely used in Bioinformatics. They are applied to protein sequence alignment, protein family annotation and gene-finding.The Baum-Welch training is an expectation-maximization algorithm for training the emission and transition probabilities of hidden Markov models. For very long training sequence, even the most efficient algorithms are memory-consuming. In this paper we discuss different approaches to decrease the memory use and compare the performance of different algorithms. In addition, we propose a bidirection algorithm with linear memory. We apply this algorithm to simulated data of protein profile to analyze the strength and weakness of the algorithm.
Shokofeh Zeinodini, Ahmad Parsian,
Volume 4, Issue 2 (3-2011)
Abstract
In this paper, a class of generalized Bayes Minimax estimators of the mean vector of a normal distribution with unknown positive definite covariance matrix is obtained under the sum of squared error loss function. It is shown that this class is an extension of the class obtained by Lin and Tasi (1973).
Abdollah Safari, Ali Sharifi, Hamid Pezeshk, Peyman Nickchi, Sayed-Amir Marashi, Changiz Eslahchi,
Volume 6, Issue 2 (2-2013)
Abstract
There are several methods for inference about gene networks, but there are few cases in which the historical information have been considered. In this research we deal with Bayesian inference on gene network. We apply a Bayesian framework to use the available information. Assuming a proper prior distribution and taking the dependency of parameters into account, we seek a model to obtain promising results. We also deal with the hyper parameter estimation. Two methods are considered. The results will be compared by the use of a simulation based on Gibbs sampler. The strengths and weaknesses of each method are briefly mentioned.
Mahdi Roozbeh, Morteza Amini,
Volume 13, Issue 2 (2-2020)
Abstract
In many fields such as econometrics, psychology, social sciences, medical sciences, engineering, etc., we face with multicollinearity among the explanatory variables and the existence of outliers in data. In such situations, the ordinary least-squares estimator leads to an inaccurate estimate. The robust methods are used to handle the outliers. Also, to overcome multicollinearity ridge estimators are suggested. On the other hand, when the error terms are heteroscedastic or correlated, the generalized least squares method is used. In this paper, a fast algorithm for computation of the feasible generalized least trimmed squares ridge estimator in a semiparametric regression model is proposed and then, the performance of the proposed estimators is examined through a Monte Carlo simulation study and a real data set.
Zahra Rezaei Ghahroodi, Hasan Ranji, Alireza Rezaei,
Volume 15, Issue 1 (9-2021)
Abstract
In most surveys, the occupation and job-industry related questions are asked through open-ended questions, and the coding of this information into thousands of categories is done manually. This is very time consuming and costly. Given the requirement of modernizing the statistical system of countries, it is necessary to use statistical learning methods in official statistics for primary and secondary data analysis. Statistical learning classification methods are also useful in the process of producing official statistics. The purpose of this article is to code some statistical processes using statistical learning methods and familiarize executive managers about the possibility of using statistical learning methods in the production of official statistics. Two applications of classification statistical learning methods, including automatic coding of economic activities and open-ended coding of statistical centers questionnaires using four iterative methods, are investigated. The studied methods include duplication, support vector machine (SVM) with multi-level aggregation methods, a combination of the duplication method and SVM, and the nearest neighbor method.
Dr Zahra Rezaei Ghahroodi, Zhina Aghamohamadi,
Volume 16, Issue 1 (9-2022)
Abstract
With the advent of big data in the last two decades, in order to exploit and use this type of data, the need to integrate databases for building a stronger evidence base for policy and service development is felt more than ever. Therefore, familiarity with the methodology of data linkage as one of the methods of data integration and the use of machine learning methods to facilitate the process of recording records is essential. In this paper, in addition to introducing the record linkage process and some related methods, machine learning algorithms are required to increase the speed of database integration, reduce costs and improve record linkage performance. In this paper, two databases of the Statistical Center of Iran and Social Security Organization are linked.
Mr Reza Zabihi Moghadam, Dr Masoud Yarmohammadi, Dr Hossein Hassani, Dr Parviz Nasiri,
Volume 16, Issue 2 (3-2023)
Abstract
The Singular Spectrum Analysis (SSA) method is a powerful non-parametric method in the field of time series analysis and has been considered due to its features such as no need to stationarity assumptions or a limit on the number of collected observations. The main purpose of the SSA method is to decompose time series into interpretable components such as trend, oscillating component, and unstructured noise. In recent years, continuous efforts have been made by researchers in various fields of research to improve this method, especially in the field of time series prediction. In this paper, a new method for improving the prediction of singular spectrum analysis using Kalman filter algorithm in structural models is introduced. Then, the performance of this method and some generalized methods of SSA are compared with the basic SSA using the root mean square error criterion. For this comparison, simulated data from structural models and real data of gas consumption in the UK have been used. The results of this study show that the newly introduced method is more accurate than other methods.
Alireza Movaffaghi Ardestani, Dr. Zahra Rezaei Ghahroodi,
Volume 17, Issue 1 (9-2023)
Abstract
Today, with the increasing access to administrative databases and the high volume of data registered in organizations, the traditional methods of data collection and analysis are not effective due to the response burden. Accordingly, the transition from traditional survey methods to modern methods of data collection and analysis with the register-based statistics approach has received more and more attention from statistical data analysts. In register-based methods, it is especially important to create an integrated database by linking database records of different organizations. Many record linkage algorithms have been developed using the Fellegi and Sunter model. The Fellegi-Sunter model does not leverage information contained in field values and does not care about specific possible values of a string variable (more common and less common values). In this article, a method that can be able to infuse these differences in specific possible values of a string variable in the Fellegi-Sunter model is presented. On the other, the model proposed by Fellegi-Sunter, as well as the method for adjusting the matching weights in the frequency-based record linkage, binding in this paper, are based on the assumption of conditional independence. In some applications of record linkage, this assumption is not met in agreement or disagreement of common variables which are used for matching. One solution used in such a case is to use log-linear model which allows interactions between matching variables in the model.
In this article, we deal with two generalizations of Fellegi-Sunter model, one with the correction of the matching weights and the other with using a log-linear model with interactions in absence of conditional independence. The proposed methods are implemented on labour force data set of Statistical Centre of Iran using R.
Mehrdad Ghaderi, Zahra Rezaei Ghahroodi, Mina Gandomi,
Volume 19, Issue 1 (9-2025)
Abstract
Researchers often face the problem of how to address missing data. Multiple imputation by chained equations is one of the most common methods for imputation. In theory, any imputation model can be used to predict the missing values. However, if the predictive models are incorrect, it can lead to biased estimates and invalid inferences. One of the latest solutions for dealing with missing data is machine learning methods and the SuperMICE method. In this paper, We present a set of simulations indicating that this approach produces final parameter estimates with lower bias and better coverage than other commonly used imputation methods. Also, implementing some machine learning methods and an ensemble algorithm, SuperMICE, on the data of the Industrial establishment survey is discussed, in which the imputation of different variables in the data co-occurs. Also, the evaluation of various methods is discussed, and the method that has better performance than the other methods is introduced.
|
|
