|
Abstract
|
The subject of "Linear Regression Analysis" is one of the important and useful subjects in diffenet fields of science including statistics. In statistics, direct methods are used to solve systems determining regression coefficients.
As there are lots of predictor variables in practical problems, if we want to determine regression coefficients with existing statistical methods, due to interdependence of some predictor variables, we have to eliminate some of these variables, losing a lot of characteristics and gathered information, as well as spending too much time for solving these kinds of problems. Therefore, it is important to introduce a method which can solve these kinds of problems without needing to eliminate the predictor variables and spending too much time. To solve these kinds of problems, in this thesis, we will discuss the iterative methods of "Block Least Squares" and "Global Least Squares".
Presenting some numerical experiments for determining regression coefficients, we will also compare the existing statistical methods with iterative methods and examine the efficiency of iterative methods in this regard.
|