For the validation sample thing, one possibility is to partition your original data file (assuming it is large) into two subsets: Say 80% for the training subset
and 20% for the validation subset. You could use the random selection capability in SPSS to create those subsets. Then, you could use the training subset to build the model and use the validation subset to validate it. The validation subset is new data because
it has not been used for building the model(s).
You may want to fix the random seed at a particular value and provide it in the manuscript in case anybody is interested in replicating your results.
Hongwei “Patrick” Yang, University of Kentucky
P.S.: I do not see this validation thing done very often in published studies. But I could be wrong.
From: SPSSX(r) Discussion [mailto:[hidden email]]
On Behalf Of stace swayne
Sent: Wednesday, October 09, 2013 8:21 PM
To: [hidden email]
Subject: Question about structural equation modeling
I received a revise and resubmit on a manuscript focused on the establishment of a measurement model for a depression scale. One of
the reviewers asked why I didn't use a validation sample. My question is, is it always necessary to have a validation sample. All opinions are welcomed.
| Free forum by Nabble | Edit this page |