Links to specific topics

(See also under "Labels" at the bottom-left area of this blog)
[ Welcome post ] [ Installation issues ] [ WarpPLS.com ] [ Posts with YouTube links ]

Friday, December 4, 2009

Welcome to the WarpPLS blog!


WarpPLS is a powerful new structural equation modeling (SEM) software. WarpPLS is commercialized by ScriptWarp Systems: www.scriptwarp.com.

Among other things, WarpPLS identifies nonlinear (or “warped”, hence the name of the software) relationships among latent variables and corrects the values of path coefficients accordingly. WarpPLS is arguably the first SEM software to do this.

Since most relationships between numeric variables are nonlinear, one could argue that WarpPLS finds the "real" relationships between latent variables in an SEM analysis. Typically path coefficients are increased, in some cases going from non-significant to significant at the P lower than 1 percent level.

The underlying algorithm employed by WarpPLS is partial least squares (PLS) regression, whose main characteristic is its ability to minimize multicollinearity among latent variables (even in the presence of overlapping manifest variables, or indicators).

Additionally, WarpPLS offers the following features, which are largely absent from most, if not all, PLS-based SEM software packages available today:
  • It estimates P values for path coefficients automatically, instead of providing only standard errors or T values, and leaving the user to figure out what the corresponding P values are.
  • It estimates several model fit indices, which have been designed to be meaningful in the context of PLS-based SEM analyses.
  • It automatically builds the indicators’ product structure underlying moderating relationships, and goes a little further. It shows those moderating relationships, related path coefficients, and related P values in a model graph as they should be shown – that is, as links between latent variables and direct links. The latter connect pairs of latent variables, while the former connect latent variables and direct links between pairs of latent variables.
  • It allows users to view scatter plots of each of the relationships among latent variables (when they are connected through arrows in the model), together with the regression curves that best approximate those relationships, and save those plots as .jpg files for inclusion in research reports.
  • It calculates variance inflation factor (VIF) coefficients for latent variable predictors associated with each latent variable criterion. This allows users to check whether some predictors should be removed due to multicolinearity (this feature is particularly useful with latent variables that are measured based on only 1 or a few indicators).

These are only a few of the new features offered by WarpPLS.


Ned Kock
WarpPLS developer

37 comments:

Sanjit said...

Hi Ned, Is WarpPLS accepted in the academic community. I mean by the management researchers. Please let me know at roysanjit2004@yahoo.co.in.

Thanks
Sanjit

Sanjit said...

Is the WarpPLS free trial version fully functional.....

Sanjit

Ned Kock said...

Hi Sanjit.

WarpPLS is currently being used by over 1,000 researchers in 33 different countries. This includes quite a few management researchers, including some with multiple publications in the very top journals in management (e.g., AMR, AMJ, ASQ, MS).

The free trial version is the full version of WarpPLS, not a demo.

Jaime León said...

Hi Ned,

First of all, I want congratulations for the easy use of the sofware.

Also, in my field, psychology, second order factor are very common, can this be done with WarpPLS? If it does, how can I do it?

Cheers

Ned Kock said...

Hi Jaime, thanks.

Yes, second order LVs can be implemented in WarpPLS. Here is a post on this:

http://warppls.blogspot.com/2010/06/using-second-order-latent-variables-in.html

Jaime León said...
This comment has been removed by the author.
Jaime León said...

Thank you very much.

Jaime León said...

And if not all of the variables belong to a second order, should I create a new data set with the value of:

- new values of latent variables
- standarized or unstandarized data?

Ned Kock said...

Your should include those values in Step 2. It doesn't matter if the data is standardized or not; the software will standardize everything for the SEM analysis.

Jaime León said...

Hi Ned,

I´ve record a video of how and what I´ve done

If it´s wrong, please, tell me where is my mistake and if right it might help somebody (like your helpful videos).

http://www.youtube.com/watch?v=xSii6VoFp9Y

http://www.youtube.com/watch?v=FinfqSG4j5w

Cheers

Ned Kock said...

Hi Jaime.

Those YouTube videos are very good. And, yes, you did everything right. I linked them into a new post, which I have just added to the blog:

http://warppls.blogspot.com/2010/06/second-order-latent-variables-in.html

Thanks!

Jaime León said...

Great you liked them, if you want (when I have some time) I could subtitled and explaien them.

Cheers

JB said...

Ned,
Is there a way to cut and paste the SEM path model with loadings and p values to include it in a document?
Thanks,
JB

Ned Kock said...

Hi JB.

WarpPLS doesn't offer the "saving to .jpg" option for the model, as it does for the plots.

The reasons is that many people prefer to draw the model themselves for inclusion in reports. That is usually what folks do when they show the results of SEM analysis in reports.

For cutting and pasting, I suggest copying the screen into a picture editor, and saving the area containing the model as a .jpg or other generic picture file. You should then be able to include it in a report.

Scott MacLean said...

Hi there. I really like this package, but I am finding it a bit buggy to install. I got the dreaded R6034 error on my Windows7 installation at home, and the suggested Path fix didn't help. However it did install OK on an XP computer. Then, at another location, it again gave the R6034 error, this time on an XP SP3 computer. I realise this is most probably to do with the Matlab side of things, but it is very frustrating. Is there any other fix you can suggest? Thanks a bunch.

Scott MacLean said...

Oops, I should also have said that the Matlab routine seems to insist on the computer's language setting being English (USA), which doesn't really suit most of the world. Again, more of a Matlab issue ?

Ned Kock said...

Hi Scott.

Yes, those problems and English requirement are related to the MATLAB Compiler Runtime (MCR).

Using the MCR leads to a tradeoff between the power of the matrix algebra enabled by it (by MATLAB, actually) and some of the MCR's installation limitations.

Interestingly, the installation problem seems to happen with only a few users. The majority don't have problems installing. For example, I never had a single installation problem. And I installed WarpPLS in quite a few machines.

In some cases the problem was clearly caused by the users not having full administrator rights on the PCs while installing WarpPLS.

Hopefully these problems will be solved in future versions, as new versions of the MCR will be used as well. Version 2.0 of WarpPLS is planed for 2011.

Anonymous said...

Hi, is there a *COMPLETE* list of features, indexes, outputs, etc. avaliable somewhere? I mean the real stats, not just marketing blurb ;)
Thanks!

Ned Kock said...

Hi Anon.

The User Manual contains a rather extensive discussion of WarpPLS's features, with various examples. It is available from WarpPLS.com, together with other items (e.g., YouTube videos with examples).

Anonymous said...

Is there a future version planned
so I wont have to purchase Matlab?

Thanks,
Mike

Ned Kock said...

Hi Mike.

You don't need to purchase MATLAB to use WarpPLS.

A new version (2.0) should be released around the middle of this year.

AndrewG said...

Hi Ned,
Congratulations on an easy to use and very effective program. I have 1 question and 1 comment.
1. VIF. My VIF results using WarpPLS showed only 1 of 6 formative LVs > 3.0. I deleted 6 of the highest cross-loading indicators in that LV, of 12 in total, saved it and ran the SEM analysis again, but there was no change in the VIF for that LV. Can you please explain why there is no change?

2. Non-linear. The warpPLS analysis also showed all 6 LVs are warped (non-linear). However, its worth commenting I think, that I notice from the warp diagrams that one or two outliers are warping the curve at each end, of each of the 6 LVs. Utilising the jacknifing setting keeps all of the coefficient paths at the same value, but lowers all except 2 p-values significantly.
Andrew

Ned Kock said...

Hi Andrew, thanks.

That can easily happen with formative LVs, because the indicators measure different facets of the LV and are not redundant. Indicators of reflective variables are redundant, and thus lead to more meaningful (or interpretable) loadings and cross-loadings after rotation.

Why do you have so many formative LVs?

Let us say you can create a formative LV with 12 indicators. That is not going to give you as clear a view of a phenomenon as two reflective LVs, one each created with 3 indicators. This will occur even though you’ll be using half of the indicators in the analysis.

Formative LVs should be using sparingly, and with caution.

Ned Kock said...

Regarding outliers influencing a curve; it is not a problem if the outliers are not due to measurement error. Some researchers want to always remove outliers from an analysis, which may be a mistake. If outliers exist and are not due to measurement error, they may provide very useful data about a phenomenon. Their influencing the curve may be a very good thing.

If you suspect that the outliers are due to measurement error, I suggest removing them from the dataset and re-doing the analysis. You can do that by saving the factor (or LV) scores into a file and then adding them to the original dataset. They will be in the exact same order as the original values. You will then be able to identify the offending cases (rows) based on the LV scores associated with the outliers and eliminate them.

Usually the outliers refer to LV scores that 2 or more standard deviations from the mean. So they are easy to spot on a dataset, especially if you order the dataset by a specific LV.

AndrewG said...

Thanks Ned. That's very helpful. With regard to the high number of formative LVs, in my field of marketing, these have been previously mis-specified as reflective.

rejiekm said...

Dear Ned,
I have two doubts regarding Interpretation of analysis results
1. What is the difference between indicator weights and indicator loadings
2.I second order formative construct with five first order constructs.I used factor scores for analyzing second order.In final analysis for path significance do i have to include first order and second order constructs or only the second order construct made of factor score .Pl help
Rejikumar,Phd scholar, India
rejiekm@rediffmail.com

Ned Kock said...

Hi rejiekm.

The User Manual explains and exemplifies the differences between formative and reflective LVs. See also this post:

http://warppls.blogspot.com/2010/01/reflective-and-formative-latent.html

I am not sure I understand you question regarding 2nd order LVs. I suggest you take a look at the post below, and if you still have questions, please elaborate more so that I can better understand them.

http://warppls.blogspot.com/2010/06/using-second-order-latent-variables-in.html

Antonio Tavares said...

Hi Ned,

My question is about dealing with outliers ... and import data to a project with a sem model already defined

Problem: removing a outlier

- I did whats in the manual ... step3 > save standardized pre-processed data > change data in excel

however, one of the consequences is that the sem model disappears, making this process very laborious

Additional question:

What I really want is to use the same model over different datasets ... saving the project with the sem model and import a new dataset to this copy without lost the sem model

Best Regards, Tavares

Ned Kock said...

Hi Antonio.

The data checking process that is conducted through steps 2 and 3 is critical in ensuring that no problems occur in the SEM analysis. This is why, whenever the dataset is changed, those steps must be conducted again.

Thanks for the input nevertheless. Easier handling of outliers is definitely something that should be considered for future versions.

Aline R. said...

Hi Ned, I got a problem when I was installing the warpPLS 2. The instalation was good, but when I tried to on it, it requested the mclmcrrt7_14.ddl. What can I do? Can you let me know at alinears@gmail.com too?
Thanks
Aline

Ned Kock said...

Hi Aline.

You should soon receive an email from the ScriptWarp Systems Support Team.

This is not a common problem.

Ned

amfeadan said...
This comment has been removed by the author.
amfeadan said...

Hi, Ned. Congratulations for this software!
I was reading the Rosipal's paper “Non linear Partial Least Squares: an Overview”. Then, I have had a doubt: the WarpPLS seeks and models non-linear relationships only among the indicators (observed variables)? Or it also seeks and models the non-linear relationship among latent variables through the predicted variable?
Thanks,
Vitor Vieira Vasconcelos
Federal University of Ouro Preto - Brazil

Ned Kock said...

Hi Vitor, thanks. WarpPLS only models nonlinear relationships between LVs. It first calculates the LV scores using a standard PLS regression algorithm, and then takes nonlinearity into consideration in the calculation of betas for the LVs that are connected by arrows.

amfeadan said...

Thanks, Ned!
Just one more question: How can I quote the WarpPLS algorithm in a scientific paper?

Ned Kock said...
This comment has been removed by the author.
Ned Kock said...

Hi Vitor. I suggest you reference the following article, which is available as a full-text PDF from WarpPLS.com:

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.

The discussion of the algorithm is under the section “Warping from Conceptual Perspective” on the paper above.