IRAQI JOURNAL OF STATISTICAL SCIENCES

In this paper,a new extended (CG) algorithms is proposed. It is in fact, a particular type of the Conjugate Gradient (CG) method which employs non-quadratic rational model, and based on inexact line searches. The Flecther and Reeves restarting criterion was employed to the standard and New versions and gave dramatic savings in computational time. The new algorithms is were promising in general, seven non linear tests function with different versions were used.

A condensed study was done to compare between the ordinary estimators. In particular the maximum likelihood estimator and the robust estimator, to estimate the parameters of the bivariate mixed model of order one, namely BARMA(1,1).
Simulation experiments were done for varieties of BARMA(1,1), using small, moderate and large sample sizes, where some new results were obtained. MAPE was used as a statistical criterion for comparison.

Classification is an important field with many applications. In particular, the classification of digital imagery has important applications in the mapping community. In this paper the comparison of three different classification methods on LANDSAT imagery of Erbil City - IRAQ: neural networks, nearest-neighbor, and discriminant analysis are made.
Out of the three approaches, k-nearest neighbors performed the best; next in accuracy was neural networks, and then discriminant analysis.

ABSTRACT
This paper includes a simulation study of three well-known filters in real lift application. These filters are - filter, Kalman filter and FIR filter. Noisily realizations were generated from three stochastic models and depending on three different families of probability distribution of noise. The modern computer application MATLAB 6 is used . The simulation results indicated that FIR filter is more powerful than the other two filters.

This research is focusing on methods related to smoothing Nonparametric Regression Functions. This is for the purpose of producing the best methods convenient for various methods. Thus, the most important purpose of the research is to find what the studies so far have offered in the field of Nonparametric Regression. Also to find alternative or modified methods, which are reliable for the treatment of failure regarding the methods in use, for example for the treatment of failure for penalized shrinkage method which depends on truncated Polynomial Regression spline model, especially for special models ,we suggest modification of knots make method more efficiency.
A simulation model has been performed with different distributions, for a number of methods. To verify the performance of such methods, many criteria have been carried out.

Forecasting of future behavior of time series is one of the important subjects in statistical sciences, because of its important need in different areas of life. Interest in forecasting had increased in the recent years, and some new techniques in the field of computer science, like Artificial Neural Networks (ANN), had appeared. These techniques have the ability learning and self- adaptation with any model, and don’t need assumptions on the nature of time series. On the other hand the classical forecasting methods, like Box-Jenkins method, need hard conditions. Hence, the need for a comparison between classical methods in forecasting of time series with ANN technique arose.

In this paper we study the migration of population among the three Iraqi cities: Nineva , Saladean, Tameem by considering these this migrations as a Markov chain with three states which represent the transitions among these cities .We obtain a transition matrix (3x3) to represent these states and we obtain that this chain is Ergodic and we find the mean and the variance of the waiting time ,the time between consecutive transition and the duration of an excursion at each state.

The objective of this research is to arrive at to the following
- Determine the number of factors that are extraced after using factor analysis for different dry milk type .
- Determine number of different Clustering groups .
- Explain the relationship between the elements after using the two analysis together .
The research depends on (20) samples drown randomly from deferent dry milks . Using the two analysis Factor and Cluster analysis and by using statistical software (SPSS), and the results are:
First: Cluster analysis
There are two factors which explain 91.8% of the total of variation between variables was used in this research . The first eigen value was 52.3% of the total variation , while the second eigen value was 38.7% of the total variation .
Second: Factor analysis
Three groups appear as a result of cluster analysis using data of (20) samples of dry milk types , as follows :
A- The first group involves full cream milk .
B- The second group involves also full cream milk ,
C- The third group involves infant formula .
Third : The two analysis together
Design figure (2) which consists of two dimensions depending on figure (1). Matching the results of the two analysis, where the cluster analysis classifies the different samples of dry milk types into three groups (as mention above). while the factor analysis mentions that the two groups (first and second) waswere two closet and they are far away from the third group . this reflects the nature of the structure of these samples, where the two groups(first and second) involve full cream milks , while the third group involves infant formula .

Most of the time series that appear in many economical geophysical and otherphenomenas are driven by non-Gaussian random error (at), so the aim of this paper is to investigate some of the probabilistic properties of Gaussian and non-Gaussian mixed model ARMA(1,1), and the identification methods of this model.
The researchers have theoretically derived the characteristic function the first four moments and the skeweness and Kurtosis coeficients for (at) for Gaussian distribution and non-Gaussian distribution (poisson) and simulation experiment were used to confirm the accuracy of the theoretical results, We have also declared the identification sample autocorrelation function (ESACF) and (Kumar) method (c-table) depending upon the pade approximation and we have suggested a method depending upon the extended sample partial autocorrelation function (ESPACF) to ascertain the efficiency of suggested method.

In this paper the radial basis function method is used for the generation of models of time series. It has been arrived at a model for nonlinear time series. It has also been found that the error is restricted between zero and one. The probability characteristics have been explained for such series accompanied by a complete analysis for the residuals of model. A simulation has been done for the purpose of examination the generative information of the model. which have the same information characteristics.

The development of life has made tons of documented data available and with the wide working of banks administration, the method of recognizing character is found.

This paper deals with the problem of estimating parameters of spatial quadratic model by Bayesian technique. This technique involves the prior information of the first and second moment of the parameters. This estimation model is called the Bayesian quadratic unbiased estimator, which is linear in the parameters. The results of estimation are compared with the estimates of minimum norm quadratic unbiased estimators and the results are encouraging.
All algorithms of computation are written by using MatLAB programming.