Links to specific topics

Saturday, June 18, 2011

Multi-group analysis with WarpPLS: Comparing path coefficients for two or more group samples


Starting in version 6.0 of WarpPLS, the menu options “Explore multi-group analyses” and “Explore measurement invariance” allow you to conduct analyses where the data is segmented in various groups, all possible combinations of pairs of groups are generated, and each pair of groups is compared. In multi-group analyses normally path coefficients are compared, whereas in measurement invariance assessment the foci of comparison are loadings and/or weights. The grouping variables can be unstandardized indicators, standardized indicators, and labels. These types of analyzes can  also be conducted via the new menu option “Explore full latent growth”, which presents several advantages (as discussed in the WarpPLS User Manual).

Related YouTube videos:

Explore Multi-Group Analyses in WarpPLS

Explore Measurement Invariance in WarpPLS

The text below is for an older version of this post. It discusses other approaches to multi-group analysis that are only partially automated, and that are thus more time-consuming to implement.

***

I previously discussed on this post multi-group analysis with WarpPLS from the perspective of comparing means of two or more groups. This is also explored in an article titled: Using WarpPLS in e-collaboration studies: What if I have only one group and one condition?

A different type of multi-group analysis would be one in which the same model is analyzed for two or more different samples, where each sample refers to a data group.

For example, a researcher could test the same model with data from the USA and Mexico. In this case, two project files would be used, and the goal of the multi-group analysis would be to assess whether the path coefficients differ significantly across groups.

An approach to conduct this type of multi-group analysis, employing the pooled and Satterthwaite standard error methods, is discussed in a recent publication (Kock, 2014). This approach is implemented through an Excel spreadsheet available from the “Resources” area of the WarpPLS.com site, under “Excel files”.

The Excel spreadsheet above takes as inputs coefficients generated by WarpPLS, including path coefficients and their standard errors. The outputs are T and P values for each pair of coefficients being compared. The formulas used in it are discussed in a recent publication (Kock, 2014).

Reference

Kock, N. (2014). Advanced mediating effects tests, multi-group analyses, and measurement model assessments in PLS-based SEM. International Journal of e-Collaboration, 10(3), 1-13.

14 comments:

Ruchi said...

In Sample 1-
The path A1(0.25) is significant (Construct a->b i..e variable
"a" effects variable "b")

Sample 2-
Path A2(0.15) is insignificant (Construct a->b i..e variable "a"
does not effect variable "b")

Combined Sample (1+2)
The path A (0.22) is significant (Construct a->b i..e variable "a"
effects variable "b")

MGA is done to find whether this difference is statically significant or not.

Now when Multigroup analysis is done between sample 1 and sample
2, the difference is "non-significant".
We say that the path coefficent A1 of sample 1 and A2 sample
2 are not statistically different (i..e difference between the two groups regarding this particular path is not significant)

BUT, what about that factor?, will the variable "a" exists for sample 2 as one of the factor effecting "b"( because in sample 1 , it is significant which means that A1=0.25) OR

variable "a" does not exists for sample 2 as one of the factor effecting "b".( because in sample 2 , it is non significant which means that A2=0)

Ned Kock said...

Hi Ruchi. I am not sure I understand the question. Having said that, please keep in mind that the approach implemented through the Excel spreadsheet takes into consideration differences between the samples – e.g., sample sizes and standard errors.

Ruchi said...

I am not able to interpret after doing multigroup analysis. I was trying in smartPLS.

If in one subsample , path is significant and in other subsample the same path is insignificant and during MGA(applying chin formula to see whether difference is significant), I found nonsignificant.

Whereas the path is significant in the total sample.

Now how to write/Interpret in the report ?

Ned Kock said...

This is an issue that requires an entire article, but in simple terms it could be stated as follows: in a multi-group analysis, one needs to ensure that the measurement models being compared are equivalent, so that one can properly compare the structural models. The equivalence and comparison are based on the measurement (or outer) and structural (or inner) model coefficients.

Here is a blog post on multi-group analysis:

http://warppls.blogspot.com/2011/06/multi-group-analysis-with-warppls.html

Under “Publications” on the WarpPLS site (www.warppls.com), you can find the doctoral dissertation by Dr. Murad Moqbel, who is a former student of mine. We made sure that he conducted a measurement model equivalence test to set the stage for a multi-group analysis targeting the structural model coefficient. The full reference is:

Moqbel, M. (2012). The effect of the use of social networking sites in the workplace on job performance. Laredo, TX: Texas A&M International University.

By the way, I think that Murad was the first to do this type of test in the context of a multi-group analysis. It entailed a lot of extra work, but I think it was worth it.

Yacine Haddoud said...

Does our data have to be normally distributed to run Keil et al's (2000) Test?

Many Thanks for your support

Ned Kock said...

Hi Yacine. I suggest employing one of the formulas in the linked pub., which is now:

Kock, N. (2014). Advanced mediating effects tests, multi-group analyses, and measurement
model assessments in PLS-based SEM. International Journal of e-Collaboration, 10(1), 1-13.

The Satterthwaite method makes fewer assumptions than the pooled standard error method.

The equation in Keil et al.'s article has a mistake. Use the article above.

Yacine Haddoud said...

Hello Prof. Kock,

Thank you so much for answering my question, In fact, I used the formula proposed in the paper you have suggested and employed Satterthwaite method. Would this require the normality distribution of data? as my data do violate this assumption!

Many Thanks Prof. Kock for your valuable support

Ned Kock said...

Since the baseline coefficients are PLS-derived, data normality is not assumed in those methods either.

Praveena said...

is it possible to do nested model comparison in warp pls? i.e, i need to do a model comparison for two models of which one is a subset of the other.

thanks
Praveena

Praveena said...

is it possible to do nested model comparison in warp pls? i.e, i need to do a model comparison for two models of which one is a subset of the other.

thanks
Praveena

ScriptWarp Systems said...

Hi Praveena. I hope that the materials linked below can be of use in connection with this.

Video: Create and Use Second Order Latent Variables in WarpPLS

http://youtu.be/bkO6YoRK8Zg

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.

http://www.scriptwarp.com/warppls/pubs/Kock_2011_IJeC_WarpPLSEcollab3.pdf

Kock, N. (2014). Advanced mediating effects tests, multi-group analyses, and measurement model assessments in PLS-based SEM. International Journal of e-Collaboration, 10(3), 1-13.

http://www.scriptwarp.com/warppls/pubs/Kock_2014_UseSEsESsLoadsWeightsSEM.pdf

Jiyoung Park said...
This comment has been removed by the author.
Caroline sujana said...

prof.... i have downloaded excel to calculate pooled standard error method. But i have 3 group...A, B, C. How to test it. is that any significant different from 3 group... cause the formula and spreadsheet only compare 2 groups. thank you.....

Ned Kock said...

Hi Caroline. Why don't you develop a table summarizing pairwise comparisons?