To provide context:

My model is a latent model; 3 Latent Variables, 3 Indicators for each of two Latent Variables and the last Latent Variable with 4 indicators.

Hence, Total of 3 Latent and 10 Measured Variables.

Paramaratization nature of CFA model:

Two latent variables are correlated with each other, while the same last latent variable is negatively correlated both of the other two.

All indicator / measured variables are expected to positively estimate unto their respective latent variables.

My Questions is

As an amateur, how to I determine / interpret / visualize relevant indices that will judge an appropriate sample size?

I never use R, but I am willing to learn through the coding. They seems okay but of course not so for a first-timer. It feels so close to finding the right spot but I am not there yet.

For clarifications on my research model, you may refer to this site: https://www.researchgate.net/publication/319433479_User's_Guide_for_the_Expectancy-Value-Cost_Survey_of_Student_Motivation

I also found a lot of resources to look at, but I am not sure which is appropriate for my project inquiry:



My model seems very close to this below, except that I have one latent variable negatively correlated with the other two: https://github.com/simsem/simsem/wiki/Example-2:-Covariance-Matrix-Specification

The matching coding pattern seems to fit those of continuousPower function of simsem: https://github.com/cran/simsem/blob/204201dd4227a6df6199a1d4fb51e017c15de77c/man/continuousPower.Rd

However, I am not sure how to implement with all the unfamiliar coding.

  • 1
    $\begingroup$ simsem makes life easier when you're running simulations. It's not a bad idea to understand simulations first, then when you understand what you're doing with simulations, you can use simsem to do them. $\endgroup$ – Jeremy Miles Sep 23 at 22:48
  • $\begingroup$ Here's an article that might help: statmodel.com/bmuthen/articles/Article_096.pdf $\endgroup$ – Jeremy Miles Sep 23 at 22:49
  • $\begingroup$ Hi, Jeremy. Thanks for the comment. I am aware of this article. However, the Mplus syntax is much more straightforward than R, which is unfortunate for free user like me. $\endgroup$ – Glenn98 Sep 24 at 4:22
  • $\begingroup$ Do you mean lavaan or simsem? I rarely use simSem (or similar things). I'd rather write it myself, and then I know what is happening. Write the model in lavaan, constrain all parameters, then get the implied covariance matrix, and use the MASS::mvrnorm() function to generate data and fit it to the model. $\endgroup$ – Jeremy Miles Sep 25 at 15:36
  • $\begingroup$ Incidentally, questions about how to do a specific thing in a specific package are considered off-topic for CrossValidated, so it's likely that your question will be closed. $\endgroup$ – Jeremy Miles Sep 25 at 15:37