CODEWITHSUNDEEP
rstructural-equation-model
HHKK
HHKK

Reputation: 21

Consideration of violation of multivariate normal distribution is not necessary in parcelAllocation (semTools) function?

I have two latent variables that were measured with parcels and one latent variable that was measured with items.

The latent variable with items did not have a multivariate normal distribution of the indicators.

When I integrate a robust estimate (e.g. estimator = MLM) in the sem.mi function to which I pass my list of 100 parcel datasets, the following is displayed:

fit.parcelsCC <- sem.mi(modCC_parcel, data = dataList, estimator = "mlm")

Robust corrections are made by pooling the naive chi-squared statistic across 100 imputations for which the model converged, then applying the average (across imputations) scaling factor to that pooled value. To instead pool the robust test statistics, set test = "D2" and pool.robust = TRUE.

I understand this to mean that no extra correction is needed, since this happens anyway by pooling the chi-squared values.

Is this correct?

Upvotes: 0

Views: 36

Answers (1)

Terrence
Terrence

Reputation: 1250

The message tells you how the test statistics were pooled across imputations, and how to request a different method. No, you don't need to do anything else.

Upvotes: 0

Related Questions