Hello everyone,
I have a question regarding the comparison of the aggregate solution to the solutions obtained via FIMIX. My question basically is if poor results for different FIMIX solutions (low EM values between 0.14 and 0.48; low variances below 0.35) actually confirm that the sample is absent of observed heterogeneity and that therefore the aggregate solution can be more confidently generalized?
For a few segments negative values of path coefficients showed up (always the smallest segments). If I stick to my aggregate solution alone can I ignore these solutions with negative values and discard them as unreliable or should I provide explanation, nevertheless?
I have a quite small sample of 70 observations and I have run FIMIX for 2 up to 5 segments. I have 13 indicators and 4 latent variables. Is maybe my sample too small to carry out FIMIX analysis in the first place?
Thank you in advance for your comments and suggestions.
Regards,
Tanja