Can someone offer an example of where a reflective construct is used as a dependent variable in a model?
In particular, my confusion is over whether each of the indicators can be taken as an equally 'important' in defining the construct? In a reflective model, does the latent variable (the construct) need to have a mathematical formula defining it, i.e. weighing the indicators?
Dear Haris,
In the way I think people usually use these terms, a reflective model is one where the lines of influence run from the concept of interest to the measures. So
(construct -> A B C) is reflective and (construct
Dear Haris, I think that you are differentiating between reflective and formative measurement models. Is this correct? A reflective measurement model means that the covariances between the indicators are used to define the latent construct. The loadings of the indicators on the latent variable may differ, therefore the indicators vary in their importance to measure the construct. A formative measurement model implies that the indicators forming the composite (which is different from a latent construct measured by reflective indicators) are a weighted linear combination of the manifest variables.
In a reflective measurement model, all indicators are important because all contribute to the latent construct. For all indicators you will get measurement equations, such as:
X1 = l*F + e, where l is the factor loading, F the latent construct (factor) and e the error term. Of course you could use F as the dependent variable in the model.
Does this help?
Regards, Karin
Dear Haris
I think Karen and Bruce have answered your question,
Based on my shallow knowledge, these terms, reflective and formative are used in the context of Partial Least Square (PLS). For us to proceed with our analysis, we must first go through the measurement model.
There are 2 types of measurement; Formative and Reflective.
Reflective measurement is the direction of causality that is FROM construct to measure. And the indicators are expected to be correlated. Therefore, we say loading.
Whereas formative measure is the direction of causality from measure TO construct. Therefore, indicators are not expected to be correlated. Hence, cannot drop an indicator from the formative measurement model.
2 articles here that do explain about reflective and formative. Hope it helps.
Best regards
Mei Peng
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling