In millenium bug resolution, till now much of focus has been upon scanning source programs to identify where to change code and upon expanding date fields in records. What has been overlooked is how to reduce tedious and expensive testing procedure. If testing shortened, it would be most advantageous because in general testing consumes more than half expense of overall developing procedure. Programs that contain Y2K bugs are mainly using elementary data types provided by programming languages. Most programs use basic calculation and I/O functionality, rather than have complex context or algorithm. Considering these characteristics, it is required to contrive more efficient test-data generation method in order to automate and improve Y2K testing. In this thesis, we propose a test-data generation method for millenium bug resolution, called domain reduction with error frequency, DOREF for short, and evaluate its correctness by simulation. Although DOREF cannot give great performance enhancement when there are many non-year-related errors, it can create more reliable test-data sets when year-related error proportion is rather high. Thus DOREF can cover the weakness of existing test-data generation method in millenium bug resolution environment. The peculiar aspect of DOREF is not only analytical determining procedure of domain-splitting points, but also introducing multiple domain-splitting points considering characteristics of variables.