[R-sig-ME] help for HLM
Douglas Bates
bates at stat.wisc.edu
Wed Sep 5 21:39:59 CEST 2012
The important part of this output is the line Number of groups: 2
You are trying to estimate three variance-covariance parameters with
only two levels of Trip. You would need many levels of Trip to be
able to do so.
When you have only two levels of a categorical variable you must model
it with fixed-effects parameters even though from the structure of the
experiment it may seem reasonable to use random effects.
On Wed, Sep 5, 2012 at 1:45 PM, Karina Villegas <villegaskary at gmail.com> wrote:
> *Dear R experts:*
> *
> *
> *I am running R version 2.12.1 on Windows 2007. I am studying the effects
> maternal behavior in the body condition of sea lion pups from California. *
>
> *
> *
>
> *I'm trying to make a hierarchical linear model*
>
> *When I run the full model if I get these results:*
>
> *> LevelModel7 <- lme(PBC ~ Sex*Dur.nurse + Sex*Freq.interaction +
> Sex*Density.females , random=~Sex|Trip, data=Dataset)*
>
> *> summary (LevelModel7)*
>
> *Linear mixed-effects model fit by REML*
>
> * Data: Dataset*
>
> * AIC BIC logLik*
>
> * 446.1461 478.1074 -211.0730*
>
> * *
>
> *Random effects:*
>
> * Formula: ~Sex | Trip*
>
> * Structure: General positive-definite, Log-Cholesky parametrization*
>
> * StdDev Corr *
>
> *(Intercept) 0.5250613 (Intr)*
>
> *Sex 0.0614132 0 *
>
> *Residual 1.4865472 *
>
> * *
>
> *Fixed effects: PBC ~ Sex * Dur.nurse + Sex * Freq.interaction + Sex *
> Density.females*
>
> * Value Std.Error DF t-value p-value*
>
> *(Intercept) 3.32002 13.521223 105 0.2455415 0.8065*
>
> *Sex 0.76586 1.584268 105 0.4834144 0.6298*
>
> *Dur.nurse 1.95957 1.798351 105 1.0896502 0.2784*
>
> *Freq.interaction -0.22624 0.219637 105 -1.0300505 0.3054*
>
> *Density.females -32.19456 19.335512 105 -1.6650480 0.0989*
>
> *Sex:Dur.nurse -0.21788 0.209796 105 -1.0385351 0.3014*
>
> *Sex:Freq.interaction 0.02165 0.025693 105 0.8426169 0.4014*
>
> *Sex:Density.females 3.15918 2.258348 105 1.3988881 0.1648*
>
> * Correlation:*
>
> * (Intr) Sex Dr.nrs Frq.nt Dnsty. Sx:Dr. Sx:Fr.*
>
> *Sex -0.998 *
>
> *Dur.nurse -0.928 0.924 *
>
> *Freq.interaction -0.194 0.195 0.103 *
>
> *Density.females 0.613 -0.609 -0.809 -0.346 *
>
> *Sex:Dur.nurse 0.928 -0.928 -0.996 -0.109 0.802 *
>
> *Sex:Freq.interaction 0.196 -0.198 -0.108 -0.994 0.353 0.108 *
>
> *Sex:Density.females -0.611 0.609 0.802 0.354 -0.991 -0.807 -0.350*
>
> * *
>
> *Standardized Within-Group Residuals:*
>
> * Min Q1 Med Q3 Max*
>
> *-2.46806088 -0.47865795 -0.05134942 0.57369682 2.64445772*
>
> * *
>
> *Number of Observations: 114*
>
> *Number of Groups: 2*
>
> * *
>
> *When I run simple models (one variable included: females Density,
> Frequency interaction) I have no problem either.*
>
> * *
>
> *> LevelModel3 <- lme(PBC ~ Sex*Density.females , random=~Sex|Trip,
> data=Dataset)*
>
> *> summary (LevelModel3)*
>
> *Linear mixed-effects model fit by REML*
>
> * Data: Dataset*
>
> * AIC BIC logLik*
>
> * 428.5119 450.1158 -206.2560*
>
> * *
>
> *Random effects:*
>
> * Formula: ~Sex | Trip*
>
> * Structure: General positive-definite, Log-Cholesky parametrization*
>
> * **StdDev Corr *
>
> *(Intercept) 1.076517e+00 (Intr)*
>
> *Sex 5.258193e-05 0 *
>
> *Residual 1.489254e+00 *
>
> * *
>
> *Fixed effects: PBC ~ Sex * Density.females*
>
> * Value Std.Error DF t-value p-value*
>
> *(Intercept) 14.774095 4.876298 109 3.029777 0.0031*
>
> *Sex -0.646479 0.566328 109 -1.141528 0.2562*
>
> *Density.females -18.980461 10.099637 109 -1.879321 0.0629*
>
> *Sex:Density.females 1.892986 1.185115 109 1.597302 0.1131*
>
> * Correlation:*
>
> * (Intr) Sex Dnsty.*
>
> *Sex -0.984 *
>
> *Density.females -0.857 0.862 *
>
> *Sex:Density.females 0.855 -0.869 -0.995*
>
> * *
>
> *Standardized Within-Group Residuals:*
>
> * Min Q1 Med Q3 Max*
>
> *-2.678669640 -0.591855777 0.006108111 0.543417969 2.424555266*
>
> * *
>
> *Number of Observations: 114*
>
> *Number of Groups: 2*
>
>
>
> *> LevelModel4 <- lme(PBC ~ Sex*Freq.interaction , random=~Sex|Trip,
> data=Dataset)*
>
> *> summary (LevelModel4)*
>
> *Linear mixed-effects model fit by REML*
>
> * Data: Dataset*
>
> * AIC BIC logLik*
>
> * 451.1833 472.7872 -217.5917*
>
> * *
>
> *Random effects:*
>
> * Formula: ~Sex | Trip*
>
> * Structure: General positive-definite, Log-Cholesky parametrization*
>
> * **StdDev Corr *
>
> *(Intercept) 1.934101e+00 (Intr)*
>
> *Sex 3.442453e-05 0 *
>
> *Residual 1.527767e+00 *
>
> * *
>
> *Fixed effects: PBC ~ Sex * Freq.interaction*
>
> * Value Std.Error DF t-value p-value*
>
> *(Intercept) 12.595151 4.211533 109 2.990633 0.0034*
>
> *Sex -0.535742 0.465667 109 -1.150484 0.2525*
>
> *Freq.interaction -0.334393 0.199095 109 -1.679564 0.0959*
>
> *Sex:Freq.interaction 0.039834 0.023241 109 1.713935 0.0894*
>
> * Correlation:*
>
> * (Intr) Sex Frq.nt*
>
> *Sex -0.943 *
>
> *Freq.interaction -0.745 0.783 *
>
> *Sex:Freq.interaction 0.743 -0.788 -0.996*
>
> * *
>
> *Standardized Within-Group Residuals:*
>
> * Min Q1 Med Q3 Max*
>
> *-2.79224714 -0.46765869 -0.04400343 0.66192444 2.53486123*
>
> * *
>
> *Number of Observations: 114*
>
> *Number of Groups: 2*
>
> * *
>
> * *
>
> *However, when I include only variable Dur.nurse, I get the following error
> message:*
>
>> LevelModel2 <- lme(PBC ~ Sex*Dur.nurse, random=~Sex|Trip, data=Dataset)
>
> Error in lme.formula(PBC ~ Sex * Dur.nurse, random = ~Sex | Trip, data =
> Dataset) :
>
> nlminb problem, convergence error code = 1
>
> message = iteration limit reached without convergence (9)
>
>
>
> *We thought it was a problem with the number of iterations and I increased
> the iterations, but I still get the error:*
>
>> LevelModel1 <- lme(PBC ~ Sex*Dur.nurse, random=~Sex|Trip, data=Dataset,
> control=lmeControl(maxIter=200))
>
> Error in lme.formula(PBC ~ Sex * Dur.nurse, random = ~Sex | Trip, data =
> Dataset, :
>
> nlminb problem, convergence error code = 1
>
> message = iteration limit reached without convergence (9)
>
>
>
> *Somebody has any idea?*
>
>
>
> *Thanks and regards*
>
> *Karina*
>
> --
> Biol. Karina Villegas Cervantes
> Estudiante de Maestría PCMyL - UNAM
>
> Laboratorio de Ecologia de Pinnipedos Burney J. Le Boueuf.
> CICIMAR-IPN
> Av. Instituto Politecnico Nacional s/n.Col.Playa Palo de Santa Rita
> La Paz Baja California Sur, Mexico.
>
> [[alternative HTML version deleted]]
>
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
More information about the R-sig-mixed-models
mailing list