Forecasting Using Regression Time Series Models

الموضوعات

إحصاء

الملخص

In classical linear regression models, we typically assume that the errors are uncorrelated, normally distributed, and having zero mean and constant variance. According to the Gauss-Markov Theorem, the Ordinary Least Squares (OLS) estimation provides the Best Linear Unbiased Estimator (BLUE) of . On the other hand, if we use the OLS estimation when the assumption of independence of error terms is violated, then the standard OLS variance estimator will be biased estimator, consequently, the usual inference procedures based on T and F tests are no longer appropriate and our inferences about the parameters estimates will be incorrect. In time series regression models where the dependent variable and independent variables are time series the assumption of independence of error terms are often violated. In this study we presented one of the most important problems that affects the accuracy of standard error of the parameters estimates of the linear regression models, which is autocorrelation problem (violating the assumption of independence of error terms in the regression models ) and we applied some statistical tests in detecting this problem such as Durbin-Watson test and Autocorrelation function (ACF) test, and for remedy we used time series regression models with ARIMA processes for errors, which is equivalent to identifying appropriate nested regression model for the time series that takes into account the autocorrelation problem in the regression model's errors and using this technique for forecasting.For practical case, we used a quarterly financial time series dataset, that represents the size of net direct credit facilities - in US dollar currency - offered by the Bank of Palestine LTD to its clients from December 2004 till March 2017 as a dependent variable and time as an independent variable.The most important findings in this study are that the autocorrelation in the regression model's errors follows ARIMA(2,0,0) model and both the Durbin- Watson test and Autocorrelation function (ACF) test are important tests for detection of Autocorrelation problem, while for remedy and elimination of this problem, time series regression models with ARIMA errors should be used, and the best nested linear time series regression model was obtained for the natural logarithms of the data among all other nested models as for dealing with autocorrelation problem and forecasting of the future observations of the natural logarithm of the series of the net direct credit facilities in the dollar currency that offered by Bank Of Palestine LTD to its clients. This result has been supported by the AIC, AICc and RMSE criteria and the residuals analysis for the nested model. Finally we used the results for forecasting future observations for the next ten quarters of the size of net direct credit facilities in the dollar currency that offered by Bank Of Palestine LTD to its clients.

روابط وملفات

التعريف والنوع

رقم الوثيقة

faec6520-1510-49a2-8277-f9b6a669ca65

رقم العقد

0

نوع الوسائط

Crawler

نوع المحتوى

الرسائل العلمية

صيغة المصدر

رسائل ماجيستير

نوع الملف

pdf text

أسماء الملفات

2103031_1.pdf

بيانات النشر

ترجمة العنوان

التنبؤ باستخدام نماذج انحدار السلاسل الزمنية

ألقاب المؤلفين

[{"name_ar":"Mahmoud K. Okasha","title_ar":"اشراف","title_en":"Supervision"},{"name_ar":"Mohammad Nabil Almasri","title_ar":"اعداد","title_en":"Preparation"}]

اللغة

English

المصدر والدورية

اسم المصدر

Forecasting Using Regression Time Series Models

المحتوى والصفحات

عدد الصفحات

0

إشراف وإعداد

الإشراف

Mahmoud K. Okasha

الإعداد

Mohammad Nabil Almasri

الاقتباسات الببليوغرافية