رفع كفاءة العلاقة المقدرة بين الدخل القومي و قيمة الصادرات الكلية المصرية

Abstract EN

One of the most prominent errors that occur when studying and analyzing time series is the problem of autocorrelation between residuals as an important regression problem.

This means that the estimated value of the error in a given time period is correlated with the estimated value of the error in a previous time period.

When studying economic time series involves a first-order autocorrelation more than a second-degree.

although it is possible that there is a negative autocorrelation that results in a error in estimated which greatly affects the results obtained and the recommendations based on them.

Therefore.

the research aims to study the nature of the problem of autocorrelation between residuals.

its causes and effects.

when studying the relationship between the value of total exports and gross national income during the period (1990 - 2017) and then apply the most important methods of detecting autocorrelation in the data to remedy this problem and finding statistical estimators with high efficiency and tested statistically after getting rid of this problem with clarification of different statistical remedy methods.

Autocorrelation is due to several reasons including: extended effects of time series data.

omission of some independent variables.

data processing.

misdeterminating of the mathematical form of the model.

Autocorrelation is detected using several methods and tests such as: graph method.

Durban Watson test.

Durban statistic test (B).

BG test Breusch-Godfrey.

"Box-Pierce-Ljung Test".

can also remedy the problem of autocorrelation in several ways.

including: transform method.

the generalized difference method.

the first difference method.

the lag of the dependent variable method.

The most important results were as follows: - Increasing the value of total exports by one unit (one million dollars) leads to increasing the national income by about 0.025 billion dollars.

- The value of the Durban Watson coefficient.

estimated at 0.345.

and the statistical significance of Q at different gaps.

as well as the significance of Breusch-Godfrey test.

estimated at 18.05.

showed a problem of autocorrelation between the residuals.

- After the general difference method (one difference) and a linear regression procedure.

the value of the Durban Watson coefficient of 0.633.

the statistical significance of Q at different gaps.

and the Breusch-Godfrey test of 16.73 (the value of the autocorrelation coefficient was 0.89).

showed a problem of autocorrelation between the residuals.

- After the general difference method (two difference) and linear regression.

it was found that the value of the Durban coefficient was estimated at 1.91.

the statistically insignificant Q at different gaps.

as well as the lack of significance of the Breusch-Godfrey test.

estimated at 2.66 (the autocorrelation coefficient was about 0.70).

showed a nonexistence autocorrelation between residuals.

- After the first difference method (one difference) and a linear regression procedure.

the value of the Durban Watson coefficient.

estimated at 0.374.

and the statistical significance of Q at some different gaps.

as well as the significance of the BreuschGodfrey test.

estimated at 21.48.

showed a problem of autocorrelation between the residuals.

- After the first difference method (two teams) and a linear regression procedure.

it was found that the value of Durban Watson coefficient.

estimated at about 2.12.

and statistically insignificant Q at different gaps.

as well as the lack of significance of the Breusch-Godfrey test.

estimated at 0.53.

showed a nonexistence autocorrelation between residuals.

- After the introduction of the dependent and independent variable with a single slowing period and a linear regression procedure.

the value of the Durban Watson coefficient.

estimated at 1.24.

and statistically insignificant Q at most of the various gaps.

as well as the lack of significance of the Breusch-Godfrey test.

estimated at 2.953.

showed no problem of autocorrelation between residuals.

- The comparison of the results showed that the presence of the autocorrelation caused many effects on the characteristics of the estimated parameters.

including: amplify the significance of the estimated parameters.

the invalidity of the use of the test t.

F.

overestimating the value of the R 2 .

Therefore.

the study recommends that researchers should pay attention to the necessary tests.