It depends on the model. One way to determine this is to do a Monte Carlo simulation. See the following paper which is available on the website:
Muthén, L.K. & Muthén, B.O. (2002). How to use a Monte Carlo study to decide on sample size and determine power. Structural Equation Modeling, 4, 599-620.
Reena posted on Wednesday, January 01, 2014 - 10:41 am
Hello. I am having trouble trying to figure out the best way to analyze my data. I have a 16 item instrument with dichotomous indicators (the indicators are observations of whether or not a task was completed by a health provider). The scale was used to observe the actions of the same health provider with one to more than one patient. I have 162 observations. I had originally drawn out a latent factor model that has 7 sub-factors that can be observed through an observation checklist. Because more missing data than I had hoped, it looks like I can only use 16 items in my analysis rather than 25. I had planned to start with CFA to test the factor model but now I think I won’t have an indicator for each of the sub-factors so I’ll have to do EFA instead. I had also planned to do structural equation modeling to look at background characteristics of the health providers and their relationship to the factors I end up with but unfortunately because of the data quality issues, I will only be able to link the health provider characteristics with about 60 observations—which I realize will not be enough.
I don’t know if my sample size is large enough even for EFA? Or maybe there’s something else I should be doing with this data? My original intention was to develop a scale (of less than 16 observable dichotomous items) to measure a latent variable. I’m trying to figure out the next best option. Thank you so much for any advice you can provide!