"It is possible to appraise the construct validity of a test interpretation without considering test use so long as no use is intended. As soon as a use is specified, then the validity investigation, including analysis of effects, must be tailored to the particular application."