ABSTRACT
To gather evidence of validity, the existence of a theory or a reasonable conceptual framework surrounding the object of measurement is of utmost importance. Three main types of evidence of observational data validity can be gathered: content validity, criterion-related validity, and construct validity. The assessment of content validity is primarily a subjective judgmental process. The Generalizability Theory and the multitrait-multimethod matrix are techniques applicable to reliability, criterion-related validity, and construct validity assessments. Nunnally suggests that factor analysis is a method that would produce evidence of content, criterion-related, as well as construct validity. For behavioral observation, there are a number of factors that can potentially reduce the validity of an observation system. These include ambiguous behavioral codes, systematic sampling error, subject reactivity, observer bias, and observer drift. Most of these threats can be minimized through systematic observer training prior to data collection, pilot observations, occasional reliability and validity checks, and careful logistic arrangements.
