[R-sig-ME] false convergence error, with bonus

[luca borger]

Hi Wade,

gald it worked. I don't remember, though, if standardizing categorical variables is similarly convenient, but probably others on the list can provide good advice - BTW, sorry, when replying I see that I had forgotten to cc the list, too.


Cheers,

Luca


-----Original Message-----
From: Wade Wall <wade.wall at gmail.com>
To: luca borger <luca.borger at cebc.cnrs.fr>
Date: Mon, 3 Jan 2011 15:29:58 -0500
Subject: Re: [R-sig-ME] false convergence error, with bonus

Hi Luca,

Thanks for the information.  Centering and standardizing the area did the
trick.  Also, thanks for the Schielzeth reference.  He sent me the paper and
it's a good read.

Do you recommend centering and standardizing categorical variables?  I
haven't ever done this, but now see that some people, including Schielzeth,
recommend it.

Thanks for any information,

Wade

On Fri, Dec 31, 2010 at 11:44 AM, luca borger <luca.borger at cebc.cnrs.fr>wrote:

> Hello,
>
> try to centre/standardize the numerical covariates, especially Area (it has
> high values, which may lead to numerical estimation problems), something
> like:
>
> snArea <- (capprod$Area - mean(capprod$Area))/(2*sd(capprod$Area))
>
> or use the scale function - I think you can do this directly within the
> lmer call (see a recent(-ish) post by Ben Bolker). See also Schielzeth,
> Holger
> (2010) Simple means to improve the interpretability of regression
> coefficients, Methods in Ecology and Evolution 1(2): 103-13.
>
> HTH and Happy New Year!
>
>
> Cheers,
>
> Luca
>
>
>
> -----Original Message-----
> From: Wade Wall <wade.wall at gmail.com>
> To: r-sig-mixed-models at r-project.org
> Date: Fri, 31 Dec 2010 11:00:03 -0500
> Subject: [R-sig-ME] false convergence error, with bonus "matrix is not
> symmetric" error
>
> Hi all,
>
> I am trying to use the lmer() function in lme4 to model counts of capsules
> on  786 plants (response variable with no zeros) spread across 23
> populations with the explanatory variables size of plant (Area), time since
> last burn of the population (TSB), and year (data collected across three
> years. Obviously, there are correlations between time since burn (TSB) and
> year, so a 2009 population that was burned in 2008 would be TSB = 1 and
> then
> in 2010 it would be TSB=3.  I tried treating year as a random factor
> (1|Year), but I got the notorious "Cholmod" error.
>
> My data set looks like this:
>
>      Site  Site_IndNum Year TSB Num_Cap       Area
> 1 PYBR002A PYBR002A_001 2008   3      18  463.34565
> 2 PYBR002A PYBR002A_002 2008   3      43  152.91702
> 3 PYBR002A PYBR002A_003 2008   3      25   80.11061
> 5 PYBR002A PYBR002A_005 2008   3      36  276.85285
> 6 PYBR002A PYBR002A_006 2008   3      39 1297.47777
> 7 PYBR002A PYBR002A_007 2008   3      23  992.42912
>
>
> cap.lme1<-glmer(Num_Cap ~ as.factor(TSB) * Area * as.factor(Year) +
> (1|Site),verbose=T,data=capprod,family=poisson,)
>
> The above model receives the error message:  In mer_finalize(ans) : false
> convergence (8)
>
> The output is:
>  0:     144803.19: 0.279342 0.100887  4.61433 0.373308 0.150254 0.000138755
> 0.000140513 4.48508e-05
>  1:     144803.19: 0.279342 0.100887  4.61433 0.373308 0.150254 0.000138755
> 0.000140513 4.48508e-05
>
>  However, when I try to remove the interaction terms, using
> cap.lme2<-glmer(Num_Cap ~ as.factor(TSB) + Area + as.factor(Year) +
> (1|Site),verbose=T,data=capprod,family=poisson,)
>
> I get the message: "Error in asMethod(object) : matrix is not symmetric
> [1,2]"
>
> Any idea what this means and how to correct it?  Needless to say, I am new
> to mixed effects modeling and my data is unbalanced.  The error seems to be
> related to Area, which is a continuous variable.  I don't get the
> nonsymmetrical matrix error message.  For example
>
> cap.lme3<-glmer(Num_Cap ~ as.factor(TSB)*
> as.factor(Year)+(1|Site)verbose=T,data=capprod,family=poisson,)
>
> works fine.
>
> I know that Area is an important variable, and so I don't want to remove
> it.
>
> Thanks for any help.
>
> Wade
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>
> __________ Information from ESET Mail Security, version of virus signature
> database 5749 (20101231) __________
>
> The message was checked by ESET Mail Security.
> http://www.eset.com
>
>
>
>
>
>
>
> __________ Information from ESET Mail Security, version of virus signature
> database 5750 (20101231) __________
>
> The message was checked by ESET Mail Security.
> http://www.eset.com
>
>
>



__________ Information from ESET Mail Security, version of virus signature database 5757 (20110103) __________

The message was checked by ESET Mail Security.
http://www.eset.com







__________ Information from ESET Mail Security, version of virus signature database 5757 (20110103) __________

The message was checked by ESET Mail Security.
http://www.eset.com