Data Extrapolation for SEM

#1
Hi,
I have a data of n=50. Is there any way in which I can extrapolate or simulate the data to n=2000, in order to carry out Structural equation modelling.
 

hlsmith

Omega Contributor
#2
Was your n=50 a random sample? You can do anything, the results are just limited. How many variables of interest are there?
 
#3
Hi, in principle you can create a bigger sample by resampling these data based on Monte-Carlo simulations (drawing with replacement). But I can't imagine that the new data either increase stability/convergence or precision of the method, because you actually do not add additional information...
 
Last edited:

spunky

Smelly poop man with doo doo pants.
#4
No. n=50 is simply too small to do anything Structural Equation Modelling related. You'll simply need to choose another method, unless you find yourself in the incredibly good position of knowing the data-generation mechanism in which case you could simulate from it.

But then again, if you knew it, you wouldn't need to do any modeling.
 
#5
Was your n=50 a random sample? You can do anything, the results are just limited. How many variables of interest are there?
I thank you very much for your earlier reply.

Let me add some points for better clarification
1. the sample is random with the assumption of normality
2. there are 5 variables and I would like to see the interaction among these 5 variables.

Looking for your reply. Thanks in anticipation