Skip to content

p-value absense in output lmer

5 messages · Karina Villegas, Søren Højsgaard, David Winsemius +2 more

Original
Søren Højsgaard
# 
The pbkrtest package provides such tests; either based on the Kenward-Rogher approximation on parametric bootstrap methods.
Regards
S?ren

________________________________________
Fra: r-sig-mixed-models-bounces at r-project.org [r-sig-mixed-models-bounces at r-project.org] På vegne af Karina Villegas [villegaskary at gmail.com]
Sendt: 16. december 2011 21:23
Til: R-SIG-Mixed-Models at r-project.org
Emne: [R-sig-ME] p-value absense in output lmer

Dear R-mixed-model-experts:

I am running R version 2.12.1 on Windows 2007. I am studying environmental
factors and maternal behavior in the body condition of sea lion pups from
California.

I specified my model as follows:

Model <- lmer
(Condicion~DuracionNurse+FrecNurse+FrecInteraccion+Temperatura+Densidad+(1|Mes)+(1|Zona)+(1|Sexo)+(1|Marea)+(1|Temporada),
family=gaussian, data=Datos)

(See output below)

At this point, my main questions are:

1. Is my model correctly built?

2. Why don?t I get p values for t?

3. Is there any way to compute p values for the fixed effects?


Linear mixed model fit by REML
Formula: Condicion ~ DuracionNurse + FrecNurse + FrecInteraccion +
Temperatura +      Densidad + (1 | Mes) + (1 | Zona) + (1 | Sexo) + (1
| Marea) +      (1 | Temporada)
   Data: Datos
   AIC   BIC logLik deviance REMLdev
 646.6 682.5 -311.3    619.1   622.6
Random effects:
 Groups    Name        Variance   Std.Dev.
 Mes       (Intercept) 2.8159e-01 5.3065e-01
 Marea     (Intercept) 2.1107e-01 4.5943e-01
 Zona      (Intercept) 2.1056e-01 4.5886e-01
 Temporada (Intercept) 1.4041e-01 3.7472e-01
 Sexo      (Intercept) 6.9748e-11 8.3515e-06
 Residual              3.8768e+00 1.9690e+00
Number of obs: 147, groups: Mes, 4; Marea, 3; Zona, 3; Temporada, 2; Sexo, 2

Fixed effects:
                 Estimate Std. Error t value
(Intercept)      69.01515   17.74172   3.890
DuracionNurse     0.04564    0.27483   0.166
FrecNurse         3.00491    0.79289   3.790
FrecInteraccion  -0.62826    0.17585  -3.573
Temperatura      -2.87229    0.91138  -3.152
Densidad        -18.78045    6.22638  -3.016

Correlation of Fixed Effects: (Intr) DrcnNr FrcNrs FrcInt Tmprtr
DuracionNrs 0.741 FrecNurse 0.945 0.582 FrecIntrccn -0.965 -0.691 -0.977
Temperatura -0.999 -0.741 -0.954 0.972 Densidad -0.919 -0.878 -0.823 0.862
0.915

I appreciate very much in advance for your help and I hope your answer.
Best Regards..

Karina


--
Biol. Karina Villegas Cervantes
Estudiante de Maestr?a 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.




--
Biol. Karina Villegas Cervantes
Estudiante de Maestr?a 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.
David Winsemius
#
On Dec 16, 2011, at 5:49 PM, S?ren H?jsgaard wrote:

            

      
      
      
    

        

      
      
    

        
Caveat: I'm not an expert in either your area or in mixed models, so  
I'm really just joining you in asking questions of the experts here.  
Q1: Does the fact that the product of the number of potential  
groupings on random effects = 4*3*3*2*2 == 144 is on the same order of  
the number of observations raise any concerns?

            

      
      
      
    

        
(DW: Edited the correlation matrix)

            

      
      
      
    

        
Q2: Does the fact that many of those numbers are above 0.95 worry  
anybody else? I would not have expected such high correlations in  
"real biological data".

  
    

      
      
    

  


  


      

    

    
1 day later

    

      
      

        
      

    

    
7 days later

    

      
      

        


  

        

      Ben Bolker
    

    

      
      #
    

  


  

      
Simon Blomberg <s.blomberg1 at ...> writes:

            

      
      
      
    

        
I believe that's correct.  I don't always centre by default,
but it's worth doing (1) for computational purposes, especially
when getting convergence warnings etc.; (2) for inferential 
purposes, especially when fitting models with interactions
(i.e. with interactions present, the main effects of parameters will
be estimated and reported at the 'zero' level of any continuous
predictors).
  So far, I haven't seen an example where the actual fitting went
wrong (silently) because of undue correlations among the input
variables.  Very large correlations (|rho|>0.99) might suggest
identifiability problems.  There is a whole literature on dealing
with collinear predictors (which is a subset of those that
can give rise to correlated parameters -- see Zuur et al 2009 
doi: 10.1111/j.2041-210X.2009.00001.x, but it's a delicate subject
(in my opinion) depending on the goal of your analysis and the
kinds of errors you're willing to subject yourself to.