Hello ResearchGate community,
Can anyone help me please with identifying the suitability of this model? Generally the ES of the relationship between latent variables should be at least moderate, even strong (especially the interrelated subscales of EIS in the center.
I have counted 64 (free) parameters... Does this seem correct? Is this model, at glance, sound?
It is estimated that we would get around 250 people for our sample, but I'm aiming to get at least 320, as according to Tabachnik and Fidell (2007) - usually 10 N per parameter is required BUT if all assumptions are satisfied (i.e. normality, linearity etc..) 5 N per parameter is permittable (hence my sample size minimum of 320, as opposed to 640).
Really appreciate any help with this... Been years since I even sniffed SEM.
What does not appear sound about this model is that the EIS factor in the middle does not have but a single latent variable (NR-Self) as an indicator. I suspect that the model in this way would be underidentified without further constraints (or additional indicators of EIS). In general, you need at least two indicators per latent variable for identification.
Is EIS supposed to be a 2nd order factor, pointing to Settled, Improved, and Harmonized? I asked because all 11 indicators of these three factors has "EIS" as the prefix. The figure title and your post also suggest that they are subscales of EIS. If they are, then the factors of these subscales are pointed by the EIS factors, no the other way round.
Shu Fai Cheung Thanks for your response - this was really helpful. I ended up abandoning all the observed variables to focus on a simple path analysis of latent variables anyway but included all for a power analysis (given I'm at planning stage this was most essential). Thanks again Shu Fai.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey