Short reports
ScienceAsia 35 (2009): 310-313 |doi:
10.2306/scienceasia1513-1874.2009.35.310
Confidence intervals for the difference between two means with missing data following a preliminary test
Pawat Paksaranuwat, Sa-aat Niwitpong*
ABSTRACT: Unit nonresponse and item nonresponse in sample surveys are a typical problem of nonresponse which can be handled by weighting adjustment and imputation methods, respectively. The objective of this study is to compare the efficiency of confidence intervals for the difference between two means when the distributions are non-normal distributed and item nonresponse occurs in the sample. The confidence intervals considered are Welch-Satterthwaite confidence interval and the adaptive interval that incorporates a preliminary test of symmetry for the underlying distributions. The adaptive confidence intervals use the Welch-Satterthwaite confidence interval if the preliminary test fails to reject symmetry for the distributions. Otherwise, the Welch-Satterthwaite confidence interval is applied to the log-transformed data, and then the interval is transformed back. Simulation studies show that the adaptive interval that incorporates the test of symmetry performs better than the Welch-Satterthwaite confidence interval when we imputed values for the missing data in two random samples based on the random hot deck imputation method.
Download PDF
19 Downloads 1180 Views
| Department of Applied Statistics, Faculty of Applied Science, King Mongkut's University of Technology North Bangkok, Bangkok 10800, Thailand |
* Corresponding author, E-mail: snw@kmutnb.ac.th
Received 2 Feb 2009, Accepted 21 Jun 2009
|
