search

cjvanlissa / tidySEM

54 stars 7 forks source link

Multigroup latent profile analysis #81

Open Gootjes opened 8 months ago

Gootjes commented 8 months ago

Hi, does mx_mixture support multigroup latent profile analysis (mplus KNOWNCLASS)

cjvanlissa commented 8 months ago

Not explicitly.. it's not clear to me exactly what KNOWNCLASS does in Mplus. But if it's a multigroup model, you might be able to build this by combining multiple OpenMx model generated by mx_mixture using the standard OpenMx procedure for multi-group models. That would probably be worth a separate paper though :)

Gootjes commented 4 months ago

Apparently multigroup LPA boils down to a variation of LTA (latent transition analysis)

cjvanlissa commented 4 months ago

Say more!

Gootjes commented 4 months ago

Christian Geiser just uploaded a new video discussing multigroup LPA https://www.youtube.com/watch?v=RYNDokqZAQY&t=210s

We "just" need to find a good example of LTA in openmx I guess :)

maugavilla commented 4 months ago

I find Mplus "language" cumbersome. I think LatentGold is more clear about this, see page 62 of the syntax guide https://www.statisticalinnovations.com/wp-content/uploads/LGSyntaxusersguide.pdf

Where we have the model parameters varied across groups, can make only class member ship varied, or also item means and variances. Seen as the LPA parameters being predicted (different) by the respective group

I think adding predictor groups for means and variances in OpenMx would not too hard, but I think its harder for class membership

maugavilla commented 4 months ago

Here is a thread of the OpenMx forum on how to add predictors in a growth mixture model, this could be useful to extend for the multiple group example

https://openmx.ssri.psu.edu/node/4428