simulateData with standardized=TRUE

[jtextor]

The function "simulateData" is currently not able to provide standardized output in the general case, as the following example demonstrates:

lav <- lavaanify( "
    y ~ (-.5) * x + (.25) * z
    x ~ .5*U
    z ~ .5*U
")
diag(cov(simulateData( lav, standardized=TRUE, sample.nobs=100000 )))

The variance of y far exceeds 1.

[ylenio]

Also, the skewness and kurtosis arguments in simulateData produce empirical skewness and kurtosis that do not always match the simulated values. For example, often empirical skewness and kurtosis values are much lower than the expected (i.e. simulated) values.

[philchalmers]

@ylenio that property may be related to this pull request: #79. But in general this data generation method isn't perfect.

[yrosseel]

5 years later, this is still not fixed. In that period, two people told me they would take this up, but nothing happened so far. It would seem I have to fix this myself.

[TDJorgensen]

I thought the simstandard package addressed this

https://cran.r-project.org/web/packages/simstandard/vignettes/simstandard_tutorial.html

[yrosseel]

Indeed. So maybe lavaan should just remove the standardized= argument in simulateData() (for now).