Use of lisrel in validating marketing constructs
There is also a need to establish that the measurement model holds across the full range of the underlying variables.If measures are found to behave differently when tested across time and across population groups, this suggests that either the measure is invalid or the concept, poorly understood.It is possible that a measurement instrument taps different constructs when applied at different times or to different population groups.Thus, repeated tests of the factorial, convergent, and discriminant validity of tests need to be made.
This means that in developing standardized instruments, correspondence rules linking theoretical dimensions to test components and subscales must be clearly established.
This can be done only if the theory is sufficiently explicit about the definition and domain of the construct.
Care must also be taken not to segregate the conceptual stage of the research from its test (Bagozzi 1980).
Another principle is that measure validation cannot be accomplished in a single study.
As in all research, apparent regularities involving measures and measurement items may be the result of chance or spurious relationships.