[R-sig-ME] Error implementing Gamma glmm with glmer

Stephani.Zador at noaa.gov Stephani.Zador at noaa.gov
Thu Dec 18 21:59:31 CET 2008


Dear all,
I am testing the influence of bottom depth and temperature on the fish catch rates (cpue). This is a 26 year time series of zero-inflated data. Therefore, I first test the presence/absence of the fish:
m1<-glmer(presence~btemp+bdepth+(1|year),data=clus1cpuebt, family=binomial, nAGQ=10)

Then, I test the non-zero catch rates, which fit the assumptions of being either lognormal or gamma-distributed:
m2<-glmer(log(cpue)~btemp+bdepth+(1|year), data=clus1subset, family=gaussian, REML=F)
m3<-glmer(cpue~btemp+bdepth+(1|year), data=clus1subset, family=Gamma(link="log"))

However, I get the error message for m3: Error in asMethod(object) : matrix is not symmetric [1,2]
Could someone please explain this error to me and how to resolve it? Further information is below:

A similar glm without the random effect is successful. 
> gmmodel4<-glm(cpue~bdepth+btemp,data=clus1subset,family=Gamma(link="log"))
> summary(gmmodel4)

Call:
glm(formula = cpue ~ bdepth + btemp, family = Gamma(link = "log"), 
    data = clus1subset)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-3.3772  -0.8814  -0.3467   0.2165   5.8240  

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept) 1.223159   0.223077   5.483 4.93e-08 ***
bdepth      0.006368   0.001638   3.887 0.000106 ***
btemp       0.343002   0.041323   8.301 2.37e-16 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

(Dispersion parameter for Gamma family taken to be 1.534698)

    Null deviance: 1636.3  on 1434  degrees of freedom
Residual deviance: 1517.3  on 1432  degrees of freedom
AIC: 11950

Number of Fisher Scoring iterations: 8

The lognormal mixed model is also successful.
> m2<-glmer(log(cpue)~bdepth+btemp+(1|year),data=clus1subset,family=gaussian,REML=F)
> print(m2,digits=4)
Linear mixed model fit by maximum likelihood 
Formula: log(cpue) ~ bdepth + btemp + (1 | year) 
   Data: clus1subset 
  AIC  BIC logLik deviance REMLdev
 4122 4148  -2056     4112    4130
Random effects:
 Groups   Name        Variance Std.Dev.
 year     (Intercept) 0.37487  0.61227 
 Residual             0.97176  0.98578 
Number of obs: 1435, groups: year, 26

Fixed effects:
             Estimate Std. Error t value
(Intercept) -1.201085   0.220070  -5.458
bdepth       0.016921   0.001313  12.885
btemp        0.533073   0.037783  14.109

Correlation of Fixed Effects:
       (Intr) bdepth
bdepth -0.595       
btemp  -0.454 -0.191

> sessionInfo()
R version 2.8.0 (2008-10-20) 
i386-pc-mingw32 

locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United States.1252;LC_MONETARY=English_United States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] lme4_0.999375-28   Matrix_0.999375-17 lattice_0.17-17   

loaded via a namespace (and not attached):
[1] grid_2.8.0  tools_2.8.0

Many thanks,
Stephani Zador, PhD
Research Associate 
Alaska Fisheries Science Center




More information about the R-sig-mixed-models mailing list