I would like to use lavaan for SEM. I am have however struggling to get things running (I went through al the examples, but they always deal with these perfect datasets). The only thing that has worked is one regression with some (five) variables as I would put into any OLS.

When I use more variables, like 10 or 15, I get the message: `Warning message:`

In lavaan::lavaan(model = ..., data = ..., model.type = "sem", :

lavaan WARNING: the optimizer warns that a solution has NOT been found!

Now, I have three questions about this:

- Does SEM indeed get into trouble when I use many variables?
- Can I circumvent this by setting starting values (for example from OLS estimates)?
- Is similarly to OLS, the outcome biased in a misspecified model?

**Contents**hide

#### Best Answer

- Does SEM indeed get into trouble when I use many variables?

It doesn't 'get into trouble' but it can have more difficulty estimating the parameters, especially if the sample size is small.

- Can I circumvent this by setting starting values (for example from OLS estimates)?

Almost certainly not. The program picks good starting values on its own, using better approaches.

- Is similarly to OLS, the outcome biased in a misspecified model?

Yes.

P.S. I have edited your question, because none of this is specific to Lavaan.

### Similar Posts:

- Solved – How to interpret this lavaan structural equation model
- Solved – lavaan works with `cfa`, provides error with lavaan(method = ‘cfa’)
- Solved – Can’t compute standard errors: the information matrix could not be inverted
- Solved – Change Score Model in lavaan
- Solved – Fitting issues when including certain combination of variables with lavaan::cfa()