sample size for sem

there is no magic number for SEM analysis. It all depends on complexity of your model and data quality (ie; loadings). Hair et al. (2010) mentioned that with good loadings even with 75 you can get a good model fit but with low loadings you may need more than 500. You may wnt to read the article by Reinartz et al. (2009) "The well-known rule of thumb that ML-based CBSEM requires at least 200 observations to avoid problems of  non-convergence and improper solutions emerged from this prior work. On the basis of our findings, we confirm that this rule is true on average but that wide variations depend on indicator loadings. Based on our analysis, the minimum sample size ranges from as low as 100 (medium or high equal loadings) to a maximum of 500 (low equal loadings and two indicators per construct). See: http://www.insead.edu/facultyresearch/research/doc.cfm?did=42571

Thanks Prof Ramayah Thurasamy! I came across with that argument as well in article written by Lei and Wu (2007) in which it does mention that the sample size required is somewhat dependent on model complexity. Thanks Prof. Now, it is clearer.