The Incidental Parameters Problem in Testing for Remaining Cross-Section Correlation