hi all,

Summary provides Std.Errors for the free cells of matrices specified in the top model.

Questions:

Is there any way to get SEs for Algebras?

Given that SEs are not obtainable currently for Algebraic matrices - how should Std.Errors - in particular the 95%CIs of a matrix - be calculated when standardizing a solution?

Do the CIs of values in a the standardized matrix retain properties as the unstandardized? For instance, that a given CI doesn't include 0?

In that case, can one obtain the SE of a standardised parameter simply by re-scaling the SE to have the same relative magnitude as held for the unstandardised parameter and SE?

So the answer to this is that we need the ML confidence interval feature.

Consider this a vote to make it one of the top priorities for the next set of feature developments.

> from Mike N

It's difficult and case-specific. It depends on partial derivatives of the function with respect to the parameters that are part of the formula for the constraints. That is why Mx uses likelihood-based CI's, and bootstrap CI's - these can be done agnostically with respect to the contributions of Newton and Leibniz. However, it may be possible for certain pathic models which have standard (ahem) formulae for standardization, to come up with SE's based on the hessian of the parameters being estimated. I used the boot function in R quite successfully (just haven't written it up as an example) - it is somewhere in the forums because I used it to check the SE's that are being reported.

It turns out that the confidence interval feature and the mixture distribution feature share some characteristics in what needs to be implemented in the backend. That's at the top of the development priority list.

fabulous