Re: what does this warnings mean? and what should I do?

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Re: what does this warnings mean? and what should I do?

Spencer Graves
          You actually received two different warning messages.  The first 8
warnings read, "optim or nlminb returned message false convergence (8)",
and the other 3 say, "... returned message singular convergence".  The
function "lmer" uses a nonlinear optimizer (either "optim" or "nlminb")
to minimize an objective function.  The first message says that the
nonlinear optimizer was still reducing the objective function when it
reached an iteration limit.  If this were the  only problem, you might
consider increasing the iteration limits, maxIter, msMaxIter, niterEM
and PQLmaxIt.  However, the "singular convergence" message says that the
estimated variance-covariance matrix of the observations became singular.

          Looking now at your output, I notice that "Corr" between
"(Intercept)" and "trust.cz1" for the "Random Effects" "commid" is
1.000.  This says that the structure of your data are not adequate to
allow you to distinguish between random effects for "(Intercept)" and
"trust.cz1" for "commid", while simultaneously estimating all the fixed
effects you have in the model.

          If I were you, I'd start be deleting all the terms from the model
that don't have a "Signif. code" beside it in the table of "Fixed
effects" and then refit the smaller model, preferably also using
'method="AGQ"'.  If I still get the same message from trying to fit the
reduced model, I would conclude that the data are not adequate to
distinguish between "(Intercept)" and "trust.cz1" for "commid".  I would
then delete "trust.cz1" from the model and go from there.

          hope this helps.
          spencer graves

ronggui wrote:

> I use lmer to fit a mixed effect model.It give some warnings.what does this warnings mean? and what should I do?
>
>
>>(fm2.mlm <- lmer(qd ~ edu + jiankang + peixun +hunyin + cadcj +
age + age2 + sex + dangyuan + Comp.1 + Comp.2+trust.cz1 +
(trust.cz1|commid), data = individual,na.action =
"na.exclude",family="quasibinomial"))

>
> Generalized linear mixed model fit using PQL
> Formula: qd ~ edu + jiankang + peixun + hunyin + cadcj + age + age2 +      sex + dangyuan + Comp.1 + Comp.2 + trust.cz1 + (trust.cz1 |      commid)
>    Data: individual
>  Family: quasibinomial(logit link)
>       AIC      BIC    logLik deviance
>  736.7059 821.8267 -349.3529 698.7059
> Random effects:
>  Groups   Name        Variance Std.Dev. Corr  
>  commid   (Intercept) 1.56413  1.25065        
>           trust.cz1   0.17922  0.42334  1.000
>  Residual             0.89728  0.94725        
> # of obs: 652, groups: commid, 39
>
> Fixed effects:
>                   Estimate  Std. Error  DF t value Pr(>|t|)  
> (Intercept)    -1.6115e-01  6.7997e-01 637 -0.2370  0.81274  
> edu            -5.2585e-02  4.1048e-02 637 -1.2810  0.20064  
> jiankang       -9.8243e-01  4.4645e-01 637 -2.2005  0.02813 *
> peixun         -4.6307e-01  2.6397e-01 637 -1.7542  0.07988 .
> hunyin         -1.2255e-02  2.8151e-01 637 -0.0435  0.96529  
> hunyin         -2.7726e-01  1.3846e+00 637 -0.2002  0.84136  
> hunyin         -2.9759e-01  8.7180e-01 637 -0.3414  0.73295  
> cadcj           2.2366e-01  7.6467e-01 637  0.2925  0.77000  
> age             9.3626e-02  4.0390e-02 637  2.3180  0.02076 *
> age2           -1.3095e-03  5.5104e-04 637 -2.3763  0.01778 *
> sex             3.9188e-01  1.9759e-01 637  1.9833  0.04776 *
> dangyuan       -5.2558e-01  5.9091e-01 637 -0.8894  0.37410  
> Comp.1          5.2463e-02  1.0309e-01 637  0.5089  0.61100  
> Comp.2         -1.5048e-01  1.1435e-01 637 -1.3160  0.18863  
> trust.cz1      -8.0709e-01  4.4632e-01 637 -1.8083  0.07103 .
> ---
> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> There were 11 warnings (use warnings() to see them)
>
>>warnings()
>
> Warning messages:
> 1: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 2: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 3: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 4: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 5: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 6: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 7: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 8: optim or nlminb returned message false convergence (8)
>  in: LMEopt(x = mer, value = cv)
> 9: optim or nlminb returned message singular convergence (7)
>  in: LMEopt(x = mer, value = cv)
> 10: optim or nlminb returned message singular convergence (7)
>  in: LMEopt(x = mer, value = cv)
> 11: optim or nlminb returned message singular convergence (7)
>  in: LMEopt(x = mer, value = cv)
>
>
>>version
>
>          _              
> platform i386-pc-mingw32
> arch     i386          
> os       mingw32        
> system   i386, mingw32  
> status                  
> major    2              
> minor    2.0            
> year     2005          
> month    10            
> day      06            
> svn rev  35749          
> language R    
>
>
>
>  
>
>
> 2005-12-14
>
> ------
> Deparment of Sociology
> Fudan University
>
> My new mail addres is [hidden email]
> Blog:http://sociology.yculblog.com
>
>
>
> ------------------------------------------------------------------------
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

--
Spencer Graves, PhD
Senior Development Engineer
PDF Solutions, Inc.
333 West San Carlos Street Suite 700
San Jose, CA 95110, USA

[hidden email]
www.pdf.com <http://www.pdf.com>
Tel:  408-938-4420
Fax: 408-280-7915

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Reply | Threaded
Open this post in threaded view
|

Re: what does this warnings mean? and what should I do?

Bert Gunter
 Spencer:

(warning: highly biased, personal opinions)

My $.02
>  Looking now at your output, I notice that "Corr" between
> "(Intercept)" and "trust.cz1" for the "Random Effects" "commid" is
> 1.000.  This says that the structure of your data are not adequate to
> allow you to distinguish between random effects for "(Intercept)" and
> "trust.cz1" for "commid", while simultaneously estimating all
> the fixed
> effects you have in the model.

Quite right. Design is the cause; overfitting/identifiability is the
symptom.
>
>  If I were you, I'd start be deleting all the terms
> from the model
> that don't have a "Signif. code" beside it in the table of "Fixed
> effects" and then refit the smaller model, preferably also using
> 'method="AGQ"'.  

Well, this might work, but it's also a prescription for overfitting a highly
biased model.

What he really needs to do is carefully rethink. What is a parsimonious
model given the data at hand? Unfortunately, this is far from a trivial
issue. Model choice for nonlinear model fitting is conceptually and
statistically difficult.

IMHO, the tendency to overfit mechanistically motivated models with
insufficient, poorly designed data is a ubiquitous scientific practice,
rarely understood by scientists (due to the complexity). As a result, there
are a lot of questionable results published in peer-reviewed literature.
Eventually it gets sorted out, but it can take a while. See Kuhn and
Feyerabend, for example.

Always enjoy your comments. Keep 'em coming.

-- Bert

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html