Need for prediction validity index

[raoul.v.kuebler]

Hallo again,

hopefully my last question:

Some PLS-Studies use Data-Splitting-Methods to gain a quality criteria for prediciton validity as the Stone-Geisser-Paradigma seems only to work on Mode-A estimations.

Am I right that I cannot just skip this value?

Does somebody know any literature beside the Chin and Todd Article from 1995 on this subject? May someone be so kindfull to explain me how to transfrom my smart pls data to get the quality criteria?

My german literature on PLS tells me to split my cases in an Estimation Sample and a Holdout Sample.

In a first step I shall evaluate the structure modell on base of the gained Estimation sample. In the second step I shall use the gained parameters in combination with the holdout sample to get the individual data for my LVs per case.

Can someone translate this for me in some PLS-for-Dummy language? May be like an IKEA-Work-Plan? So which data do i have to split? How can I evaluate the structure modell? How do I get the individual LVs for each case?

Thanks ALOT for your help and be sure you save my academic career...

Raoul