On Testing Outlying Observations