Testing Random Effects in Linear Mixed Regression Models

نخلة, ساندي صفوت فايق; الروبي, محمد جمعة; عبداللاه, ايمن  احمد امين

doi:10.21608/safq.2025.370748.1181

Testing Random Effects in Linear Mixed Regression Models

نوع المستند : المقالة الأصلية

المؤلفون

¹ كلية التجارة، جامعة السويس

² كلية التجارة - جامعة السويس

³ كلية التجارة، جامعة المنوفية

10.21608/safq.2025.370748.1181

المستخلص

Linear mixed models are extensions of regression models in which data have a hierarchical structure with units nested in clusters (Snijders and Bosker, 1999). They are multilevel models where the data are in the form of groups or clusters, Accounting for the correlation that might exist within group elements and that is why linear mixed models require a special treatment rather than the standard regression. It is essential to test for the need for the existence of the random effect in the model as it will be power-consuming to estimate an unneeded parameter, so it will be helpful to test whether the Linear mixed models are a better procedure. This article is designed to represent a comparison using a simulation study to examine the performance of five tests that aimed to test the existence of the random effect. This comparison depends on comparing those tests in different settings, varying the number of groups and number of observations among those groups, to detect the best performance under small size samples, also Varying the distributions of the error term to indicate the least affected test with the distributional assumption. The criterion of the comparison will be the power and the size of the test under these different factors.

الكلمات الرئيسية

الموضوعات الرئيسية