Comparison of Stratification Method and Its Application Using National Population Health Survey Data

• Activity Number: 688
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Sunita Ghosh* and Qiaohao Zhu
• Companies: Alberta Health Services Cancer Care and Government of Alberta
• Keywords: Stratification ; concordance threshold ; Youden Index ; c-index ; NPHS ; minimum p-value approach

Abstract

Clinical and epidemiological studies are often challenged by stratification of a continuous variable into categorical. The issues associated with the stratification are that loss in sample size and also misclassification. In clinical setting stratification helps in making clinical significance and easy interpretation. In the present study we compare the four methods: minimum p-value approach, Youden index, overall c-index and maximal concordance threshold. We hypothesize that overall c-index and Youden index will provide most unbiased estimates and accurate stratification with minimum bias. These four methods will be compared using simulated data sets and then will be applied to National Population Health Survey (NPHS) data. To validate our results we will be using two-fold cross-validation and Jackknife methods.