forum.smartpls.com

I think it would be useful to add a feature "power analysis" (e.g. reporting degrees of freedom, calculating sample sizes etc.) for a specified model.

Hi,

could you please give us more information on what you need - if you perpare a list and briefly explain how to calculate it we might be able to include your suggestions in the next release (--> okay, sample size is clear; --> df for bootstrapping...?).

Best
Christian

Hi Christian,

power analysis is crucial to calculate appropriate sample sizes for a given causal model (a priori!) and is recommended by many researchers. In fact, also Chin recommends it, since the "10 times the most complex indicator..."-rule is just a simple heuristic (Chin/Newsted 1999).

There are a lot of tools for calculating power. I'd recommend Paul Dudgeon's "niesem", since it is freeware as far as I know. See also http://rubens.its.unimelb.edu.au/~dudgeon/. However, it runs in DOS-mode. Check it out to see how it works.

One of the parameters which one has to specify is the df for the whole causal model. For the other parameters (e.g. alpha- and beta-values, power-value) one can easily check the literature on power analysis (e.g. McQuitty 2004, JBR Vol 57). Until now, SmartPLS does not report df, as far as I know.

Concluding, I don't think it is absolutely necessary to implement a complete power analysis-feature at this very moment. Reporting df for a drawn causal model will do!

Best Regards,
Samy

Hi,

I know this is an old thread, but it seems to be one of the few to deal with sample size and statistical power. I'd like to suggest having a look at GPower (http://www.psycho.uni-duesseldorf.de/aa ... ts/gpower/), which implements the power calculations suggested by Cohen (1988) for a number of typical analyses. Results are much more precise than power tables.

It's freeware, too, and although it is a DOS program an installer for Windows is available.

Just my 2 cents,

Joachim Schroer

Take a look at gpower 3 in the same website, it runs on windows and is amazing in the types of situations it can handle. The site also provides a list of referred journals containing papers using gpower.

Hello,

thank you for pointing me to the new 3.0 release (yeah, finally! :-) ) . It really is a major improvement over the previous 2.0 version and much easier to use.

Kind regards,

Joachim