HELP WITH LAVAAN ERROR. Error with path analysis using latent variable

39 views

Skip to first unread message

Lindy Howe

unread,

Jan 2, 2020, 2:07:53 PM1/2/20

to lavaan

Hi all - I am completing a mediation analysis using lavaan. The analysis has 2 causal factors (TAI and mBDI), 1 mediator (BPD), and 1 DV (SUD) which is a latent variable.

I am attempting to run this model using lavaan, but am getting an error when fitting that I cant fix!! Any help? Syntax below.

> model <- ' SUD.sum =~ Alc.Sum+Drug.Sum+MJ.Sum

+ BPD.Sum ~ TAI.SUM + mBDIII.SUM

+ SUD.sum ~ BPD.Sum + TAI.SUM + mBDIII.SUM

+ mBDIII.SUM ~~ mBDIII.SUM

+ TAI.SUM ~~ TAI.SUM

+ mBDIII.SUM ~~ TAI.SUM

+ BPD.Sum ~~ BPD.Sum

+ SUD.sum ~~ SUD.sum'

> fit <- lavaan(model, estimator = 'MLF', data=df)

Error in lav_model_estimate(lavmodel = lavmodel, lavpartable = lavpartable, :

lavaan ERROR: initial model-implied matrix (Sigma) is not positive definite;

check your model and/or starting parameters.

Note: The error goes away when I remove the latent variable, but I need the latent variable to be included.

Thanks!

Terrence Jorgensen

unread,

Jan 3, 2020, 11:16:43 AM1/3/20

to lavaan

I am attempting to run this model using lavaan

Try using sem() instead, which automatically adds the residual-variance estimates you have omitted from your syntax (the cause of your error).

I hope your N is large enough for MLF. First-order information does not perform well with modest sample sizes.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam