Modeling formative measurement models in AMOS requires the researcher to set specific values (indicator weights, error variances). How do you obtain these values when you have no reference (pretested) scale (with e.g. Cronbach's alpha) available?
Mark, you can do it. Below are articles that discuss the strategy on how to implement formative measurements in SEM:
Diamantopoulos, A., Riefler, P., & Roth, K. P. (2008). Advancing formative measurement models. Journal of Business Research, 61, 1203–1218.
Diamantopoulos, A. (2011). Incorporating formative measures into covariance-based structural equation models. MIS Quarterly, 35(2), 335–358.
Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in mearkting and consumer research. Journal of Consumer Research, 30(2), 199–218.
Petter, S., Straub, D., & Rai, A. (2007). Specifying formative constructs in information systems research. MIS Quarterly, 31(4), 623–656.
Kind of. I have three indicators/items A, B, and C, which are caused by construct X (reflective). Together with construct Y and Z, these three constructs form a second order construct (formative).
Amos requests me to fix error variances for my indicators/items and indicator weights. But I don't know what to choose here.
I am not an expert in SEM but as I knew one of the assumptions of covariance based SEM was reflective measurement model. So i don't think it can handle it. Please share with us if you know how to do it. Thanks in advance!
Mark, you can do it. Below are articles that discuss the strategy on how to implement formative measurements in SEM:
Diamantopoulos, A., Riefler, P., & Roth, K. P. (2008). Advancing formative measurement models. Journal of Business Research, 61, 1203–1218.
Diamantopoulos, A. (2011). Incorporating formative measures into covariance-based structural equation models. MIS Quarterly, 35(2), 335–358.
Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in mearkting and consumer research. Journal of Consumer Research, 30(2), 199–218.
Petter, S., Straub, D., & Rai, A. (2007). Specifying formative constructs in information systems research. MIS Quarterly, 31(4), 623–656.
it is most important to (re)consider exactly what the meaning of such a latent variable should be as the literature (especially the marketing folks around Diamantopolous) seems to view such a variable as a composite/index of the measures (which turns "measures" or "indicators" into "parts", "facets" etc.) and the ontological stance of the variable into a fiction/aggregate.
A SEM model instead which contains a latent variable and its causal "indicators" which have effects on the latent variable is a completely different issue (and these two views were mixed for decades). Here, the latent variable and its causal "indicators" are two different entities - the "indicators" have causal effects on the latent variable and the latent variable - in turn- has effects on other outcomes. As Wilcox et al. emphasize, this implies that causes and the dependent latent variable has to be distinct (and hence the indicators are no parts). This implies further that the latent variable is a variable like any other latent variable and could be MEASURED by reflective indicators (Thats why such a model is no formative *measurement* model, because causal indicators CAUSE and not measure the latent variable.
See
Howell, R. D., Breivik, E., & Wilcox, J. B. (2007). Reconsidering formative measurement. Psychological Methods, 12(2), 205-218.
Wilcox, J. B., Howell, R. D., & Breivik, E. (2008). Questions about formative measurement. Journal of Business Research, 61, 1219-1228.