Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: "not concave"


From   William Buchanan <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Re: st: "not concave"
Date   Fri, 1 Mar 2013 06:12:13 -0800

The first correction to make here is more semantic than anything. You are estimating a model with one factor and four manifest/observed variables.  Are you sure that you are estimating identical models across all programs?  Perhaps you are constraining a different path in other packages.  Which estimator did you use with Mplus (note the spelling since like Stata, Mplus is not an acronym)?  I'm not a user of anything related to SPSS, but providing your Mplus syntax could also be helpful for those of us that also use Mplus.  

Have you tried using any of the maximization options (e.g., difficult or specifying a different algorithm from the default)?  Are all of your manifest variables continuous? Were you using the same estimator in the other packages?

You've provided just enough information to create more questions for people trying to help you.  I highly recommend looking at the Statalist FAQ (as Nick has already pointed out to you) for helpful guidance creating questions that are likely to yield helpful responses from the listserv users.

HTH,
Billy

Sent from my iPhone

On Mar 1, 2013, at 5:24, Wahideh Achbari <[email protected]> wrote:

> Thanks for this Nick. Here is the syntax:
> 
> sem (Poleff -> opinion_not_vote) (Poleff -> say_pol) (Poleff ->
> care_pol) (Poleff -> pol_not_complex), method(mlmv) standardized
> latent(Poleff ) nocapslatent
> (47 all-missing observations excluded)
> 
> There was no fit at all as the iterations kept running with "not
> concave" in brackets.
> 
> If you need the data I can also send that.
> 
> Thanks again,
> 
> On 1 March 2013 14:18, Nick Cox <[email protected]> wrote:
>> It's not the model; it's your log-likelihood function that is awkward
>> over part of the parameter space.
>> 
>> It's important to realise also that maximum likelihood is emphatically
>> not an algorithm. It is an estimation method.
>> 
>> This isn't necessarily a big problem. Nor is the absence of warning
>> messages from other software necessarily diagnostic. Perhaps Stata is
>> just more explicit, or they used different algorithms to implement
>> their fits. It's difficult to say.
>> 
>> However, it is not clear from your post whether the command you used
>> -- not even spelled out here -- converged to give a fit.
>> 
>> To say more experts in this technique (not me) would need to see your
>> command syntax and results from Stata.
>> 
>> Your post is strong on "I think I have a problem" but much more
>> context is needed for a fuller answer. See the FAQ on advice on this.
>> 
>> Nick
>> 
>> On Fri, Mar 1, 2013 at 12:55 PM, Wahideh Achbari
>> <[email protected]> wrote:
>> 
>>> Here's a query about a very simple confirmatory factor analysis with 4
>>> factors. I have been trying to estimate this model, but unfortunately
>>> the model is "not concave" running a Maximum Liklihood algorithm. What
>>> surprises me most is that I have been running the same model in AMOS
>>> and MPLUS without any difficulties. I would like to know what can be
>>> done to remedy this. Your help is greatly appreciated.
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> *   http://www.ats.ucla.edu/stat/stata/
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index