AVE LOW (second order construct)

[Isaac F]

Hello,

I have a second order construct, which consists of 3 LOC with a reflective-reflective design:

1. The AVE of each of the LOCs is >0.5

2. The loadings of the indicators of each of the LOCs are >0.7

3. The loadings of the repeated indicators (the reflective indicators of the HOC) are >.7 except one which is >0.5

However the AVE of the HOC is relatively low, AVE=0.35,

Is it acceptable?

Thanks,

Izak

[Husameddin Dawoud]

You should do that manually because smartPLS calculate AVE for HOC wrongly. I don't know about warpPLS.

to do that, lets say we have three LOCs and the path coefficient between LOCs and HOC as the following: 0.75, 0.72, 0.65. Then the calculation will be like this

(0.75)^2 + (0.72)^2 + (0.65)^2 / 3 = 0.501. notice that devided by the number of LOCs which is 3.

[Ned Kock]

AVEs are calculated in WarpPLS as the average of squared loadings, as in the illustration by Husameddin.

 

Often higher order LVs will conform better to the definition of formative LVs, even when they are made up of lower order reflective LVs.

 

Formative LVs, as well as other related issues and references, are discussed in the User Manual for WarpPLS, and in the following articles, which are available from the “Publications” area on the www.warppls.com website.

 

Kock, N. (2011), Using WarpPLS in e-collaboration studies: Mediating effects, control and second order variables, and algorithm choices. International Journal of e-Collaboration, 7(3), 1-13.

 

Kock, N. (2010). Using WarpPLS in e-collaboration studies: An overview of five main analysis steps. International Journal of e-Collaboration, 6(4), 1-11.

 

There are several criteria for assessment of formative LVs starting on page 59 of the WarpPLS 4.0 User Manual.

 

Ned

--
Join us at the PLS Applications Symposium: http://plsas.net
---
You received this message because you are subscribed to the Google Groups "PLS-SEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

No virus found in this message.
Checked by AVG - www.avg.com
Version: 2014.0.4158 / Virus Database: 3629/6858 - Release Date: 11/22/13

---

No virus found in this message.
Checked by AVG - www.avg.com
Version: 2014.0.4158 / Virus Database: 3629/6859 - Release Date: 11/22/13

[Isaac F]

Thanks very much!,

 

I am using SmartPLS too. I have calculated the AVE as you have indicated, now it is 0.456. Is it too low or should I accept that? Is it because the loadings of each the reflective indicators of the HOC (repeated indicators) are between 0.4 to 0.7? If yes how should I deal with that?

 

Again, I appreciate your support, I have so many questions and I don't get much help.

--

Join us at the PLS Applications Symposium: http://plsas.net
---
You received this message because you are subscribed to the Google Groups "PLS-SEM" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+unsubscribe@googlegroups.com.
To post to this group, send email to pls-sem@googlegroups.com.

[Husameddin Dawoud]

Low factor loading causes low AVE value. Try to delete the lowest factor loading one by one until you reach to the acceptable value of AVE (>0.5). If you use repeated indicator approach then don't forget to delete the item from both HOC and LOC.

Regarding the support of smartPLS; the forum of the application http://www.smartpls.de/forum/ is very helpful. you can use search menu to check weather there is an answer to your inquiry or not. otherwise you can post here or in that forum what you want.

Regards,

[Isaac F]

Thank you;
In my study the loadings of the LOC's indicators are excellent; however the loadings of the HOC refelctive indicators are low (approximately 0.45). Should I delete these indicators from both LOC and HOC or should I use the two stage approach instead?
Best wishes
Zak

To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.

[Husameddin Dawoud]

try to delete the low factor loading from both HOC and LOC and check AVE continually.

I suggest to check cross loading first. Indicator’s  outer loading  on  the  associated  construct  should  be  greater  than  all  of  its loadings on other constructs. It should be grater by at least 0.1 more than other cross loadings (Hair et al., 2013)

[Osveh Esmaeelnejad]

Hi , I have the same question,U mean for calculating the AVE of second order ,we should consider path coefficient of the second order and first order and not the AVE of first orders? and the number two always is fix? I mean for example I have two first  orders with 0.6 and 0.53 , and and the path coefficient is .85 and 0.95 , now how could I calculate?could u explain again sorry. and is there any reference based on the formula you said?

[Ramayah T]

Hi Osveh,

I agree with the view point of Prof. Jose Roldan as he commented earlier as I also teach this method in my workshops.

T. Ramayah

From my point of view, you should calculate the second order construct's AVE by using the loadings of the first order constructs.

Best regards,

José L. Roldán

--

Join us at the PLS Applications Symposium: http://plsas.net
---
You received this message because you are subscribed to the Google Groups "PLS-SEM" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
http://www.ramayah.com/

[Christian Ringle]

Hi

Just use the AVE of the HOC as correctly reported by SmartPLS. In PLS path modeling, you use an indicator re-use technique. Hence, you assign all reflective indicators of the LOCs to the reflective HOC. The PLS path modeling algorithm computes the latent variable scores of the HOC based on these assigned indicators. Hence, their outer loadings determine the AVE of the HOC. Some of my recent articles applying (and evaluating) hierarchical component models (often called second-order models) in PLS-SEM:

Klarner, Patricia, Marko Sarstedt, Christian M. Ringle, and Michael Höck. 2013. "Disentangling the Effects of Team Competences, Team Adaptability, and Client Communication on the Performance of Management Consulting Teams." Long Range Planning 46 (3): 258-286.

Kocyigit, Orhan, and Christian M. Ringle. 2011. "The Impact of Brand Confusion on Sustainable Brand Satisfaction and Private Label Proneness: A Subtle Decay of Brand Equity." Journal of Brand Management 19 (3): 195-212.

Some recent conceptional articles:

Ringle, Christian M., Marko Sarstedt, and Detmar W. Straub. 2012. "A Critical Look at the Use of PLS-SEM in MIS Quarterly." MIS Quarterly 36 (1): iii-xiv.

Becker, Jan-Michael, Kristina Klein, and Martin Wetzels. 2012. "Hierarchical Latent Variable Models in PLS-SEM: Guidelines for Using Reflective-Formative Type Models." Long Range Planning 45 (5-6): 359-394.

And you may also want to check the section on HOC in the book on PLS-SEM:

Hair, Joseph F., G. Tomas M. Hult, Christian M. Ringle, and Marko Sarstedt. 2014. A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM). Thousand Oaks: Sage.

Best
Christian Ringle
www.tuhh.de/hrmo

[Isaac F]

Thanks for that.

However, Becker et al. (P.378) suggested that: "...What distinguishes the assessment of higher-order constructs from that of first-order constructs is the role of the weights and loadings in the analysis: they are not obtained from the relations between higher-order construct and manifest variables, but are obtained from the relations between higher-order construct and lower-order constructs. This distinction is especially important if the repeated indicator approach is used, as the weights and loadings are now represented by the path coefficients between higher-order and lower-order constructs, and not by the manifest indicators that are repeated at the construct level".

According to that, and of course, this is my interpretation of the text: the AVE should be manually computed for the second order construct by averaging the squared loadings of the second order and the first orders and not as calculated by the SmartPLS.

[S. Mostafa Rasoolimanesh]

Thank you all.

I would like to add one point to Issac's explanation. I also agree that the estimated AVE and CR by SmartPLS considers all outer loadings of HOC including loadings of repeated indicators and LOCs, so the estimated AVE and CR is not correct and should be calculated manually based on just LOC's loadings. However, if we follow this procedure, still the estimated AVE and CR isn't correct, because in this condition the loadings of repeated indicators effect on the loadings of LOCs. So if we calculate manually the AVE and CR according to LOCs' loadings, still these estimated criteria can be wrong. It seems that two- stage approach suggested by some scholars such as Hair et al., 2013 can be a better approach. This approach can be done by SmartPLS in two stages.

Pls correct me if my explanation is wrong.

Best Regards

Mostafa

[Christian Ringle]

But your interpretation means that you loose information... I would not suggest this kind of AVE manipulation.

[Christian Ringle]

Hi Mostafa,

certainly a work around.

In addition to the two stage approach that you mentioned (which we also described in the MISQ article), we commented on three pitfalls of HCM (see below). I generally suggest starting with a proper higher order component which one my want to break down into its subcomponents.

In my view, one of the best explanations of HCM and examples have been given by
Lohmöller, Jan-Bernd. 1989. Latent Variable Path Modeling with Partial Least Squares. Heidelberg: Physica.

He uses "general ability" of school children as the higher order component and, then, differentiates "spatial", "verbal" and "numerical" abilities as lower order components. In his explanations, he distinguishes between "special and general factor models" and "multiple battery model". A must read!

Best
CR

Hair, Joseph F., Christian M. Ringle, and Marko Sarstedt. 2013. "Partial Least Squares Structural Equation Modeling: Rigorous Applications, Better Results and Higher Acceptance." Long Range Planning 46 (1-2): 1-12.

"Hierarchical component models are relatively easy to conduct in PLS-SEM by using Wold’s
(1982) repeated indicator approach. Recent explications provide additional insights into
how hierarchical component models should be incorporated into PLS path models (Becker
et al., 2012; Hair et al., 2013; Ringle et al., 2012; Wetzels et al., 2009). Unfortunately, users often
misapply higher order models both conceptually and technically. Instead of presenting
proper explications that justify the type of higher order model used (e.g., the reflectivereflective
type or the reflective-formative type), researchers routinely use hierarchical component
models to summarize information in a second, third or even higher dimension of abstraction
with the plausible objective of reducing the number of relationships in the PLS path
model. The resulting higher order models however are often difficult to defend from a theoretical
point of view.
Despite its ease of application, the repeated indicator approach may entail several pitfalls. First,
the number of indicators per lower order component should be balanced (Becker et al., 2012).
Otherwise the estimated relationships between the higher and the lower order components may
be biased. Second, researchers usually do not evaluate the higher order constructs, although the
same evaluation criteria (and their critical values) used for the lower order components apply.
Hence, information about relevant evaluation criteria outcomes is important and should be
provided. Third, some researchers include relationships from other latent variables in the structural
model, which are not part of the hierarchical component model, to formative higher order
constructs. These relationships always have a value of approximately zero when the indicator
reuse technique is applied to determine the higher order construct in PLS-SEM, because the
formative lower order components already explain all of the former’s variance. Hence, the conclusion
that other constructs in the structural model do not explain any variation of the
reflective-formative type or the formative-formative type higher order construct would be false
and misleading, as this is a technical outcome of the repeated indicator approach. In such situations,
a two-stage approach (Ringle et al., 2012) should be used, which allows for handling
this technical limitation of the repeated indicator approach."

Best
CR

[Isaac F]

Thank you very much.
The detailed  explantion and the references you have provided to Mostafa are highly useful. I feel now that I can properly justify the approach has been taken to assess the reliability and validity of the HOC.

[S. Mostafa Rasoolimanesh]

Thank you very much Prof. Ringle for your reply and explanation.

Best Regards

Mostafa

[777299598]

I hope all of you are fine.I have the same problem of AVE of the higher order construct. I used the latent scores of the first order constructs and the AVE became more than the cutoff .50. Is my test right and is this the tow step approach. I mean after I used the the repeated approach I have a low AVE for the heir order construct.Then I replaced the indicator of the second order construct by the latent scores of the lower order constructs which first computed in pls and saved then I use them as indicators for the second order construct >Thus the resulted model includes first order latent constructs measured by 23 items or indicators and the second order construct measured by five indicators which are the latent scores of the first order constructs which previously calculated>AVE became more than the cutoff .50.Am I right

[Ned Kock]

Just to be clear - you are not using WarpPLS, correct? 

I don't think you'd have this problem with WarpPLS. 

Btw, the paper below provides a good illustration of higher order LV analysis:

Schmiedel, T., vom Brocke, J., & Recker, J. (2014). Development and validation of an instrument to measure organizational cultures’ support of business process management. Information & Management, 51(1), 43-56.

http://www.scriptwarp.com/warppls/pubs/Schmiedel_etal_2014_OrgCulturesBPM.pdf

[Ned Kock]

If one uses the approach for higher order LV analysis outlined in the video below, and the second (or higher) order LV turns out to be formative, you should use criteria for formative measurement quality assessment. Judging from experience, it is quite common for second and higher order LVs to "become formative".

http://youtu.be/bkO6YoRK8Zg

These criteria (for formative measurement quality assessment) are discussed in WarpPLS' manual (link below, search for "formative" inside the document). Normally a formative LV will have are relatively low AVE, because the AVE is the average of the sum of squared loadings.

http://www.scriptwarp.com/warppls/UserManual_v_4_0.pdf

Ned

[suje...@gmail.com]

Hi,

I had the same problem of a low AVE with the higher order construct and could improve it with the way suggested by you. Thank you.

I would also like to know how to calculate the discriminate validity and composite reliability of these higher order constructs since the Smart pls Report gives values below the threshold.Pl. help me.

Best regards,

Sujeevi

[Discord]

##- Please type your reply above this line -##

Hi Pls-sem,

We just wanted to confirm that we have received your inquiry (1355767). You put time and effort into contacting us, so our Support Team will make sure we give your inquiry the attention it deserves.

Cheers,
The Discord Team

Pls-sem Aug 16, 02:53 PDT

>> 2013/11/23 Husameddin Dawoud <hda...@gmail.com>

>> >>> You should do that manually because smartPLS calculate AVE for HOC >>> wrongly. I don't know about warpPLS. >>> >>> to do that, lets say we have three LOCs and the path coefficient between >>> LOCs and HOC as the following: 0.75, 0.72, 0.65. Then the calculation will >>> be like this >>> (0.75)^2 + (0.72)^2 + (0.65)^2 / 3 = 0.501. notice that devided by the >>> number of LOCs which is 3. >>> >>> >>> On Friday, 22 November 2013 19:26:25 UTC+8, Isaac F wrote: >>>> >>>> Hello, >>>> >>>> I have a second order construct, which consists of 3 LOC with a >>>> reflective-reflective design: >>>> >>>> 1. The AVE of each of the LOCs is >0.5 >>>> >>>> 2. The loadings of the indicators of each of the LOCs are >0.7 >>>> >>>> 3. The loadings of the repeated indicators (the reflective indicators >>>> of the HOC) are >.7 except one which is >0.5 >>>> >>>> However the AVE of the HOC is relatively low, AVE=0.35, >>>> >>>> Is it acceptable? >>>> >>>> Thanks, >>>> >>>> Izak >>>> >>> -- >> Join us at the PLS Applications Symposium: http://plsas.net

>> <http://www.google.com/url?q=http%3A%2F%2Fplsas.net&sa=D&sntz=1&usg=AFQjCNGP5BTaiGxF04ytpfVnX2B4DOe-cg>

>> --- >> You received this message because you are subscribed to the Google Groups >> "PLS-SEM" group. >> To unsubscribe from this group and stop receiving emails from it, send an

>> email to pls-sem+u...@googlegroups.com. >> To post to this group, send email to pls...@googlegroups.com.

>> For more options, visit https://groups.google.com/groups/opt_out. >> >

--

Join us at the PLS Applications Symposium: http://plsas.net --- You received this message because you are subscribed to the Google Groups "PLS-SEM" group.

To unsubscribe from this group and stop receiving emails from it, send an email to pls-sem+u...@googlegroups.com. To post to this group, send email to pls...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.

[J. Henseler]

Please note that you  seem to follow outdated guidelines.

The only consistent approach for reflective higher-order constructs is described here:

 

Riel, Allard C. R. van; Henseler, Jörg; Kemény, Ildikó; Sasovova, Zuzana (2017). Estimating hierarchical constructs using consistent partial least squares: The case of second-order composites of common factors. Industrial Management & Data Systems, 117 (3), 459-477, doi:10.1108/IMDS-07-2016-0286.

 

This article is open access and  can be downloaded from http://dx.doi.org/10.1108/IMDS-07-2016-0286

 

Kind  regards,

Jörg Henseler

 

--

Prof. dr. ir. Jörg Henseler

Chair of Product-Market Relations

Head of the Department of Design, Production and Management

Faculty of Engineering Technology

University of Twente.

P.O. Box 217, 7500 AE Enschede, The Netherlands

Phone: +31 (0)53 489 2953

E-mail: j.hen...@utwente.nl

Web: http://www.henseler.com

 

Newest guidelines on PLS path modeling:

-         Henseler, J., Hubona, G., Ray, P. A. (2016). Using PLS path modeling in new technology research: updated guidelines. Industrial Management & Data Systems, 116 (1), 2-20 (click here to download).

 

Current publications:

-         Henseler, J. (2017). Bridging design and behavioral research with variance-based structural equation modeling. Journal of Advertising, 46 (1), 178-192 (click here to download).

-         de Koeijer, B.; Wever, R.; Henseler, J. (forthcoming). Realising product-packaging combinations in circular systems: Shaping the research agenda. Packaging Technology and Science, in print (click here to download).

-         van Riel, A. C. R.; Henseler, J.; Kemény, I.; Sasovova, Z. (2017). Estimating hierarchical constructs using consistent partial least squares: The case of second-order composites of common factors. Industrial Management & Data Systems, 117 (1), in print.

-         Schuberth, F.; Henseler, J.; Dijkstra, T. K. (forthcoming). Partial least squares path modeling using ordinal categorical indicators. Quality & Quantity, in print (click here to download).

[Ned Kock]

Note that you can, as a possible solution to your problem, build analytic composites (aka “indices”, but not in the model fit index sense) in WarpPLS by aggregating lower order LVs. The following video clips may be useful:

 

Create and Use Second Order Latent Variables in WarpPLS

 

http://youtu.be/bkO6YoRK8Zg

 

Explore Analytic Composites in WarpPLS

 

https://youtu.be/bxGi0OY8RD4

 

Best, Ned

---

This email has been checked for viruses by Avast antivirus software. www.avast.com

[Sujeevi Jayaratne]

Thank you Ned. I will try this out.

Best regards,

Sujeevi

--
Join us at the PLS Applications Symposium: http://plsas.net
---

You received this message because you are subscribed to a topic in the Google Groups "PLS-SEM" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/pls-sem/qwToFJR1InA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to pls-sem+unsubscribe@googlegroups.com.
To post to this group, send email to pls...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.