عناصر مشابهة
Modeling Correlated Missing Data
المصدر: | مجلة البحوث المالية والتجارية |
---|---|
الناشر: |
جامعة بورسعيد - كلية التجارة
|
المؤلف الرئيسي: | |
مؤلفين آخرين: | , |
المجلد/العدد: | ع4 |
محكمة: | نعم |
الدولة: | مصر |
التاريخ الميلادي: | 2017 |
الصفحات: | 234 - 254 |
DOI: | 10.21608/jsst.2017.60555 |
ISSN: | 2090-5327 |
رقم MD: | 1215279 |
نوع المحتوى: | بحوث ومقالات |
اللغة: | English |
قواعد المعلومات: | EcoLink |
مواضيع: | |
رابط المحتوى: |
|
المستخلص: | The missing data have occurred in longitudinal studies. Missing data may reduce the performance of confidence intervals, reduce statistical power and increase the standard errors. This paper provides a new model for arising the effect of missing data, which are considered also as correlated binary data, on the correlated binary variables. Depending on a serial dependence we formulate the model using Markov properties for the correlated missing data. The alternative quadratic exponential form is employed for the correlated binary variables. The logit functions are used for modeling the missing data. The vectorized generalized linear models are used with five cases of correlated missing data associated with covariates. The simulation study is investigated, using "bindata" and "VGAM" packages in R program, to indicate the effect of missing data on the regression model comparing with the model without missing data. |
---|