how to specify MODE A or MODE B for reflective-formative HOC with extended repeated indicators

Frequently asked questions about PLS path modeling.
Post Reply
PLS User
Posts: 11
Joined: Sun Aug 29, 2010 5:08 pm
Real name and title:

how to specify MODE A or MODE B for reflective-formative HOC with extended repeated indicators

Post by jamarin » Sun Dec 29, 2019 8:15 pm

I have a model with reflective-formative HOC and formative-formative HOC and I want to estimate it with extended repeated indicators.
I have read
Sarstedt, M., Hair, J. F., Cheah, J.-H., Becker, J.-M., & Ringle, C. M. (2019). How to specify, estimate, and validate higher-order constructs in pls-sem. Australasian Marketing Journal (AMJ), 27(3), 197-211. doi:
Becker, J. M., Klein, K., & Wetzels, M. (2012). Hierarchical latent variable models in pls-sem: Guidelines for using reflective-formative type models. Long Range Planning, 45(5-6), 359-394.

But I If have properly understand both papers, the figure 8 in the Sarstedt et al. 2019 may be wrong.
In that paper authors wrote page 200
"While researchers typically use Mode A to estimate reflectively specified measurement models and Mode B to estimate formatively specified measurement mod- els, Becker et al. (2012) show that this choice of measurement mode for repeated indicators does not apply to the orientation of the lower-order components, but the higher-order component. Specifically, their simulation study shows that Mode B estima- tion of the higher-order component in a reflective-formative type higher-order construct produces the smallest parameter estima- tion bias. Hence, even though the (repeated) indicators identifying the higher-order constructs are specified reflectively on the lower- order components, researchers should use Mode B for these re- peated indicators on the higher-order component. In light of these findings, researchers should use Mode A for a reflectively specified higher-order constructs (i.e., reflective-reflective and formative- reflective types) and Mode B for formatively specified higher- order constructs (i.e., reflective-formative and formative-formative types). This recommendation also holds for the first stage of the embedded two-stage approach. In contrast, the disjoint two-stage approach should be estimated using the standard settings on both stages; that is, Mode A for reflectively specified measurement models and Mode B for formatively specified measurement models."
In page 206 they add "In order to apply the extended repeated indicators approach, we need to extend the model by drawing direct relationships between the four antecedent constructs and the two lower-order compo- nents of reputation ( Fig. 8 ). Following Becker et al. (2012) , we use Mode B to estimate the measurement model of the higher-order component REPU . As is the case in"

But in figure 8 (representing reflective-formative HOC) , REPU construct was drawn exactly in the same way than in figure 6 (representing reflective-reflecitve HOC specification) That is, arrows from REPU to items (comp1 ..like3) that sounds as MODE A. And I think that arrows should be drawn from items (comp1 .. like3 ) to Construct (REPU).
PS. figure 7 and 9 (the version for disjoint two stage approach) are coherent and arrows are drawn in the expected direction.
Juan Marin

SmartPLS Developer
Posts: 1110
Joined: Tue Mar 28, 2006 11:09 am
Real name and title: Dr. Jan-Michael Becker

Re: how to specify MODE A or MODE B for reflective-formative HOC with extended repeated indicators

Post by jmbecker » Wed Jan 22, 2020 2:49 pm

I agree that Figure 8 is misleading and I would have liked to present it differently, but some of the co-authors think it is better this way.

The estimation Mode A and Mode B are not necessarily linked to the mode of measurement (reflective / formative). SmartPLS uses as a default Mode A for reflective and Mode B for formative, because this is generally a good choice but there might be situations where this does not apply. Thus, you can change this default setting by double clicking on the construct (or right-clicking) and change the mode of measurement. You do not necessarily need to change the direction of arrows.
Why do some co-authors think it is represented better this way: Because the indicators are still reflective on the first-order they think it would be better to also place them reflective on the higher-order construct. They are reflective indicators and not formative indicators. Ok, but I think this is still confusing. Because at the end the logic is similar to normal constructs: if the construct is formative (here the higher-order construct) then attach the indicator formatively (and use Mode B estimation), even if these are repeated indicators from some lower order constructs.

Nevertheless, as you also correctly mention, we write on page 206 that "we use Mode B to estimate the measurement model of the higher-order component REPU."
Dr. Jan-Michael Becker, University of Cologne, SmartPLS Developer
GoogleScholar: ... AAAJ&hl=de

Post Reply