[R-sig-ME] Model fit after false convergence
Luciano La Sala
lucianolasala at yahoo.com.ar
Tue Oct 27 23:32:32 CET 2009
Dear R-crew,
I am using lmer to fit a model for epidemiological data. I start from a saturated model containing all the main effects of interest (all with p < 0.25 in a bivariate screening) plus one random intercept (please see below). From there, I need to refine my model based on AIC criterion.
As you can see below, I'm getting a convergence error after fitting the saturated model, which I don't know how to deal with.
I read somewhere that this problem can be solved by adding "verbose = TRUE" in the model. After doing this, I get the output below.
At this point my questions are:
1. Are thses results worth trusting?
2. What does "verbose=T" do? If not added, the only thing I get is the "convergence error" message. If added, the model keeps running after the "convergence error" line.
3. What do the strings of numbers after "0: ... ", "1: ..." and "2: ..." below represent? I'd never seen that before.
4. The error message "Warning message: In mer_finalize(ans) : false convergence (8)" still shows up. Should I worry, or just disregard it and start simplifying my model from there on?
5. I have only 196 observations, but 104 random effects. Maybe this is the problem?
Thank you so much in advance!!
Luciano
CODE AND OUTPUT:
> full.model1 <- lmer(Death~HatchOrder + Year + ClutchSize + EggBreadth + EggVolume + ClutchVolume + I(ClutchVolume^2) + Asynchrony + SibingCompetence + (1|NestID),family=binomial,1,verbose = TRUE)
0: 210.20372: 1.18952 -4.38021 -2.14958 -2.47874 -0.643502 0.412270 0.811849 -1.48282 -0.875658 -0.0777208 -0.677601 -0.0518550 -0.604514 0.0650300 -0.000209959 1.31029 1.03084
1: 210.10114: 1.18952 -4.38021 -2.14958 -2.47874 -0.643502 0.412270 0.811849 -1.48282 -0.875658 -0.0777208 -0.677601 -0.0518550 -0.604514 0.0650299 -0.000211591 1.31029 1.03084
2: 210.10114: 1.18952 -4.38021 -2.14958 -2.47874 -0.643502 0.412270 0.811849 -1.48282 -0.875658 -0.0777208 -0.677601 -0.0518550 -0.604514 0.0650299 -0.000211591 1.31029 1.03084
Warning message: In mer_finalize(ans) : false convergence (8)
> full.model1
Generalized linear mixed model fit by the Laplace approximation
Formula: Death~HatchOrder+Year+ClutchSize+EggBreadth+EggVolume+ ClutchVolume+I(ClutchVolume^2)+Asynchrony+SiblingCompetence+(1|NestID)
Data: 1
AIC BIC logLik deviance
244.1 299.8 -105.1 210.1
Random effects:
Groups Name Variance Std.Dev.
NestID (Intercept) 1.4150 1.1895
Number of obs: 196, groups: NestID, 104
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -4.3802131 5.6178408 -0.7797 0.43557
HatchOrderSecond -2.1495845 1.2872835 -1.6699 0.09495 .
HatchOrderThird -2.4787439 1.8973058 -1.3065 0.19140
Year2007 -0.6435022 0.4793135 -1.3425 0.17942
ClutchSizeTwo-eggs 0.4122699 3.2797887 0.1257 0.89997
ClutchSizeThree-eggs 0.8118488 4.9473010 0.1641 0.86965
BreadthCATBLarge -1.4828236 0.7150390 -2.0738 0.03810 *
BreadthCATBMedium -0.8756582 0.8789277 -0.9963 0.31911
BreadthCATBSmall -0.0777208 1.0925961 -0.0711 0.94329
VolumeCATBLarge -0.6776006 0.7200754 -0.9410 0.34670
VolumeCATBMedium -0.0518550 0.9619276 -0.0539 0.95701
VolumeCATBSmall -0.6045140 1.1920193 -0.5071 0.61206
ClutchVolume 0.0650299 0.0828151 0.7852 0.43231
I(ClutchVolume^2) -0.0002116 0.0001869 -1.1318 0.25770
Asynchrony 1.3102881 0.4369016 2.9990 0.00271 **
SibCompPresent 1.0308412 1.0857487 0.9494 0.34240
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Yahoo! Cocina
Encontra las mejores recetas con Yahoo! Cocina.
http://ar.mujer.yahoo.com/cocina/
More information about the R-sig-mixed-models
mailing list