Comparison of Different Models for Predicting the Probability of Interview Completion in Complex Telephone Surveys

• Activity Number: 141
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods
• Author(s): Xian Tao *+ and Robert Montgomery and Zhen Zhao and Meena Khare
• Companies: NORC and CDC and National Center for Health Statistics
• Keywords: National Immunization Survey ; Sample Release ; Probability of Completes

Abstract

To manage large and complex survey sample, the sample is divided into batches and each batch is released based on the performance of previous batches. The accuracy of predicting the performance of the released sample affects the forecast of the future release amount. There are many factors such as call outcomes, case disposition, number of refusals and number of dials that affect the probability of a completed interview. The National Immunization Survey (NIS) is a nationwide list-assisted RDD survey that collects immunization histories of children within the age range of 19-35 months. Using data from NIS, we constructed several models, e.g. simple linear regression, logistic regression, CHAID, survival analysis and neural networks, to measure the performance of the released sample and compare the accuracy and efficiency of different prediction methods.