he term Generalized Estimating Equations indicates that an estimating equation is not the result of a likelihood-based derivation, but that it is obtained by generalizing other estimating equation. Liang and Zeger (1986), Zeger and Liang (1986) and prentice (1988) developed a most recent method of estimating the parameters of the marginal model. They present a class of estimating equations that take the correlation into account to increase the efficiency. This class of estimating equations is called Generalized Estimating Equations (GEE) and the correlation matrix is called working correlation matrix. The name working is used in the sense that it is an approximate correlation matrix of i Y 's (the response variable).The estimates of ? obtained by GEE are consistent and in addition consistent variance estimates can be obtained under the weak assumption the weighted average of the estimated correlation matrices converge to a fixed matrix.

The stock exchange data are highly correlated from one company to other company.

In our study, we have used stock exchange data because, we want to compare the different method of estimation. We have taken six company and their closing price and the general index in 2009 of DSE. In our calculation we have taken general index as an explanatory variable and the different company closing price as a independent variable. We want to estimate the following model

Selection of correlation structure is an important issue in Generalized Estimating Equation (GEE). We have earlier stated that there are four correlation structure namely, independent, exchangeable, autoregressive and pair wise. One needs to select under which correlation structure is unknown GEE works well or provides efficient estimate in several situation.

We have concentrated on standard error as a measure of accuracy for an estimator of the parameter. So in this case, we look only on the standard error under different correlation structure of different methods of estimation procedure.

In this study we consider stock exchange data, this data are highly correlated.

In the previous situation, we see that the standard error of the parameters of the GEE method is lowest than any other method. From this study we may conclude that from different method of estimation the GEE parameters are gives the efficient estimate and best approach.