Some Effects of Randomization in Hypothesis Testing with Correlated Data.
Research and development technical rept.,
ARMY ELECTRONICS COMMAND WHITE SANDS MISSILE RANGE N MEX ATMOSPHERIC SCIENCES LAB
Pagination or Media Count:
The effects of each of four different randomization schemes upon the Students-T test statistic are investigated under the assumption that the data are correlated. Two specific correlation matrices are considered. The results indicate that the statistic T may be biased either too large or too small to be reliable in hypothesis testing. Whether T is biased or unbiased depends upon the randomization scheme chosen. Author
- Statistics and Probability