How do I obtain cross loadings to assess discriminant validity of my constructs?
One way is to take your data set and use a spreadsheet such as Excel.
Standardize all the indicators you choose. This means creating new indicators
for each of the ones in your model by first calculating the means and standard
deviations of each indicator. Then subtracting the mean and dividing by the s.d.
to create the standardized variables.
Now create construct scores by using the weights PLS-Graph provides on each of
the indicators for the construct in question.
??? Do you mean to add the weights together to get the construct scores???
Yes, to create the construct scores - you multiply each standardized indicator
with their respective weight provided from pls.
Now, correlate the construct scores with the standardized indicators. If you did
it right, the correlations of the indicators to their respective construct
should be identical to the PLS-Graph output. The rest are the cross-loadings.
The other way is to have PLS-Graph calculate the construct scores and
standardized indicators as part of the output. Then cut and paste this
information to a spreadsheet or stat package and do the correlations.
????Can this be done under the "Options - Output" menus?????
Yes - under final results - go past construct scores to data matrix. It will
provide standardized (rescaled) indicators and construct scores.