I'm trying to use the simulation tool to test a model from an experiment that is hard to replicate. I fed the tool with normal variables with their means, standard deviations, and mix and max values, sourced from Analyze -> Descriptive Statistics -> Descriptives; and a correlation matrix sourced from Analyze -> Correlate -> Bivariate -> Pearson; double checked all the inputs, and the resulting correlation matrix is drastically different from the inputted one. I tested the model that started with 198 measured values, with the new 1000 generated cases, and the model fit indexes dropped and the RMSEA increased considerably... and I mean considerably.

Is this deviation with the correlation matrix an expected result, or am I doing something wrong?

I'm a complete noob with simulations, I see them as a black box, and since I'm using 58 variables, it took me ours to fill that correlation matrix, so sadly, playing around and testing different settings with the simulation tool isn't an option, and that's why I decided to post my problem in this forum, hoping that a simulation expert might be able to help.    

Hi,

Discrepancies between input and resulting correlation matrices in simulations, especially with many variables, might suggest problems with the simulation process or input data. Check the accuracy and format of your data, including means, standard deviations, and min/max values. Ensure your correlation matrix is positive-semidefinite for realistic outcomes.

Review your simulation method, as some may not preserve the correlation structure. Look into other ways of generating correlated data, like multivariate distributions or copulas. Account for model errors, possibly using the Tucker-Koopman-Linn model.

Regarding the model fit and higher RMSEA, these measures can be impacted by model size. As indicators increase, RMSEA may decrease, implying a better fit, but this isn't always so. Other factors might be causing the poor fit. Consider reevaluating your model or adding residual covariances based on modification indices.

Hope this helps.

<quillbot-extension-portal></quillbot-extension-portal>