Re: LCA: model failed to determine a valid solution...why?

This seems unusual.

With a 1-class model (and no missing data) estimation really needs just
one iteration, so convergence shouldn't be an issue there.

Suggestions:

1. Are data being read correctly? Is there a way to check--e.g., a
report of N's and means for each variable?

2. Can you run any benchmark data and verify that you get the right
results? That's a good way to make sure you've formatted data files
correctly, are using the proper options, etc.

3. Have you experimented with changing the convergence criterion
and/or maximum number of iterations?

4. If you're getting results (even though convergence isn't reached),
do they look plausible?

(This is my third attempt over a several-hour period to reply to the
post. If the reply ends up appearing multiple times, please excuse.)

--
John Uebersax, PhD

.

Relevant Pages

  • Re: can NN have variable number of features
    ... Iteration and convergence are not involved. ... i don't have missing data but they are short than the max ... imputation gets complicated. ...
    (comp.soft-sys.matlab)
  • Re: LCA: model failed to determine a valid solution...why?
    ... With a 1-class model (and no missing data) estimation only needs one ... iteration, so convergence shouldn't be an issue there. ... Have you experimented with changing the convergence criterion and/or ...
    (sci.stat.edu)
  • Re: Old chestnut: f(f(x))=exp(x)
    ... convergence*, and possibly some other things not rigorously proved!). ... we have a way of expressing the general iteration of exp- 1 as ... a power series, subject to the convergence provisos. ...
    (sci.math)
  • Re: Richardson-Lucy deconvolution ?
    ... corresponding pixel in the observed image, and p_is the ... i.e. the fraction of the pixel at j observed at i.. ... There are two widely used ways to initialize the RL iteration. ... Regarding the convergence, you could check some norm of the difference ...
    (sci.image.processing)
  • Re: can NN have variable number of features
    ... i just have training/testing ... H is decided by newrb i don't have to set it ... Iteration and convergence are not involved. ...
    (comp.soft-sys.matlab)