You know we have not any Identified method for sampling in SEM method,so can we use sampling software for example Gpower ,NCSS,IBM sample size for our sampling .if yes ,HOW ?
Hi,
I will suggest to follow the rules which we use when doing Exploratory Factor analysis. 5 as a minimum and 10 as a maximum participants for each item in the scale. As we follow EFA with CFA or SEM then we accept the sample method as described. Thus, you have to cunt how many items or parameters and multiply by 5 to get the mimimus needed sample size. If you multiple by 10 and it is feasible to get the sample, then this is better.
hope this help.
The shot answer is that you can calculated, but its depends of the complexity of the model. Sadly, Gpower its not advanced enough to give you information about the required sample size of a SEM model (in my knowledge), but you can get an approximation with this link if you know how many latent variables you want to use. http://www.danielsoper.com/statcalc3/calc.aspx?id=89
Hello Hamid,
That is a rule of thumb rather than Gpower analysis. I was taught that for SEM (applied on Mplus) minimum sample size is 200 and optimal sample size is 300.
Best wishes
I would recommend using Gpower to calculate the minimum sample size to to detect an effect.. Next, I would draw out your most complex hypothesized model to estimate how many parameters you will have. You'll need to know/decide things like if going to use latent or observed variables. To be safe, you might add a few extra parameters to prepare for the possibility of needing to modify your model (e.g., theoretical or design adjustments). This will allow you to determine a minimum sample size for the model structure. Depending on a range of factors (e.g., complexity of the study, types of measures), either of these minimums could be lower, but the key will be use the higher minimum of the two.
You might also use this site: http://www.danielsoper.com/statcalc3/calc.aspx?id=89 which will provide you with both minimums based on information you provide. They also provide formulas and references (I have not reviewed or used the site before, so do investigate them closely). If you run the default numbers (the ones pre-programmed in), you'll see that that it gives you two minimums and then recommends you use the higher.
Hi,
I will suggest to follow the rules which we use when doing Exploratory Factor analysis. 5 as a minimum and 10 as a maximum participants for each item in the scale. As we follow EFA with CFA or SEM then we accept the sample method as described. Thus, you have to cunt how many items or parameters and multiply by 5 to get the mimimus needed sample size. If you multiple by 10 and it is feasible to get the sample, then this is better.
hope this help.
Another good program that may help is Optimal Design. I think it is also free and is a little less user friendly, but more sophisticated (in my opinion at least) and has a very helpful manual. Both are great programs, but Optimal Design may be a little better suited in this case. And, the general guidelines given above of # of observations per variables are very good general guidelines.
Hello there, just in case someone is interested in NCSS 10, it is now available in cloud with 30-day free trial and great price after. take a look at https://www.apponfly.com/en/ncss10?KAI
I'm really happy to find this discussion. Thanks Jan for the excellent citations. As a quick update, the website shared by a few people ( http://www.danielsoper.com/statcalc3/calc.aspx?id=89 ) has been moved, but is still available and worth the visit ( https://www.danielsoper.com/statcalc/calculator.aspx?id=89 ). In fact, I'd suggest just bookmarking the master page here https://www.danielsoper.com/statcalc/default.aspx .
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size