|
|
Dear Everyone,
Can you suggest/refer a reference suggesting the cut-off of regression weights, particularly weights on the "indicators of latents"?
A latent which is included in a model contains more than 15 indicators (seven-point ordinal response options). Fifteen is too many for a latent, thus it would be trimmed down. I am thinking of using parceling approach, but I doubt because only one set of data is available for the analysis. One set of data meaning, I have data only for use in testing the model and I have no other data for conducting, say factor analysis or parceling, in the latent.
Your suggestion is highly solicited.
Thank you,
Eins |
|
Locked
