Share
مقدرات طريقة بيز وبعض الطرائق التقليدية شبه المعلمية لتقدير دالة الانحدار اللوجستي في ظل البيانات المفقودة == Bayesian Method Estimates And Some Semiparametric Regular Estimating Methods For Estimating The Logistic Regression With Missing Data
Author name:
علي محمد علي جيجان الخفاجي
Supervisor name:
قتيبة نبيل نايف القزاز
General topic:
Administration and Economics
Specific topic:
Statistics
Degree:
Master
University:
University of Baghdad - Faculty Of Administration And Economics - Department Of Statistics
Language:
Arabic
University location:
Baghdad
First pages:
07T3896 - p.pdf
Abstract:
يلقى موضوع تحليل النماذج شبة المعلمية والذي يدمج النماذج المعلمية والنماذج اللامعلمية اهتماما واضحا في معظم الدراسات والتي تاخذ طابعا اكثر تقدما في عملية التحليل الاحصائي الدقيق الذي يهدف الى الحصول على مقدرات ذات مستوى عال من الكفاءة , في بعض الدراسات | Semi - parametric models analysis is one of the most interesting subjects in recent studies because of the precise way it describes the statistical databy giving efficient parameters. In some studies the response variable takes two values, either zero - no response - or one - response - which is called the logistic regression model. And the observations of this model may suffer from missingness and that is when the objective of the study comes.In this thesis the researcher studied some of the methods of estimating missing observations if the missingness is in the parametric variables or in the nonparametric variables. When the missingness was in the parametric variable two methods has been used, which are the Unconditional Mean (UCM) and the KBNS methods. And when the missingness was in the nonparametric variable two methods has been used which are Unconditional Mean (UCM) and Conditional Mean (CM) methods.After completing the observations the methods ofestimating the parameters in Semi - parametric modelswere used to estimate the parameters and these methods were (NW) and (CLLE) and the Bayesian method. Then the results were compared using MSe criteria.A simulation study had been made to compare different sample sizes, variances and missingness ratios by using the MSequality criteria. And the most important conclusions were, in the parametric variable the methods gave close results while the nonparametric variable methods gave different results according to the sample sizes and the variances.Then the best methods in the simulation study with sample size 90 and variance 0.2 were applied on the heart failure disease because the real data sample size was 92 with variance 0.23 and (CLLE) method gave the best result.
