Test for significant differences between path coefficients

LenaH · Post by **LenaH** » Tue Aug 13, 2013 12:16 pm

Hello everyone,

I always find this forum extremely helpful. Thank you for the lively discussions here. Until today, I always found answers to my questions by just reading former posts. However, now I'm struggeling with one specific problem.

I want to test if two path coefficients in a single model (no multi group comparison) differ significantly from each other. I doubt that it is sufficient to just conduct a t-test. Does anyone have the same issue and was able to solve it? Are there any literature recommendations (I already searched, but not successfully)?

Thanks in advance!

christian.nitzl · Post by **christian.nitzl** » Tue Aug 13, 2013 7:45 pm

Hi,

You can use the same approach for testing the difference of two path connections in one single model as you us for path connections in different models. You can use Chin’s test or Henseler’s group test or simple the difference of the bootstrap distribution in an Excel template.

Greetings,

Christian

LenaH · Post by **LenaH** » Tue Aug 13, 2013 9:10 pm

Thank you sooo much! This problem was bothering me quite a while. I already worked with Henseler's test in a multigroup scenario.
Do you have any literature suggestion that I could quote in order to show that this approach is valid?

Thank you again!

christian.nitzl · Post by **christian.nitzl** » Thu Aug 15, 2013 9:15 am

At the moment I do not know a source for PLS-SEM who exactly matches to your question. However, because regression analysis and PLS-SEM are very similar you can use e.g. following source which refer to a regression analyses:

http://www.udel.edu/soc/faculty/parker/ ... CRIM98.pdf

But you have to pay attention that the formula for testing the difference is here a little bit different because in PLS-SEM you use standard errors from a bootstrapped distribution.

All the best,
Christian

LenaH · Post by **LenaH** » Thu Aug 15, 2013 9:18 am

Thank you so much! I have this paper right here on my desk and I'm happy that you second my thought of citing this paper.
Thanks again, your thoughts were a big help to me!

christian.nitzl · Post by **christian.nitzl** » Thu Aug 15, 2013 9:35 am

By the way, I have an Excel template which you can use for testing the difference. That template is based on the corrected formula by Chin that you can find in my working paper on page 46 (in the orignal formula was a small bug):

http://papers.ssrn.com/sol3/papers.cfm? ... id=2097324

Best regards,

LenaH · Post by **LenaH** » Thu Aug 15, 2013 9:48 am

thank you, I just downloaded the paper!