1
$\begingroup$

I posted a question recently regarding general linear mixed effects models, and I think I may have finally specified the glmer model correctly. I am interested in finding any differences in home range sizes between parks, between 2 time periods (wet and dry season). I have added individual ID as a random effect and year nested within ID as some IDs have multiple measurements for different years. I have 8 levels for "season", and 2 levels for "park". My question is: why is the 8th level for "season" omitted from the fixed estimates? Also, by looking at the p values, am I correct in stating that home range sizes for season 3 are significantly different between parks? I am definitely new to mixed models and would appreciate any help.

Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
Family: gaussian  ( log )
Formula: homerange ~ park * season + (1 | ID/year)
Control: glmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 1e+05))

 AIC      BIC   logLik deviance df.resid 
2271.9   2332.7  -1117.0   2233.9      162 

Scaled residuals: 
Min      1Q  Median      3Q     Max 
-3.7332 -0.4334 -0.0220  0.3912  4.1532 

Random effects:
Groups   Name        Variance Std.Dev.
year:ID  (Intercept)  275.2   16.59   
ID       (Intercept)  123.1   11.10   
Residual             1997.7   44.70   
Number of obs: 181, groups:  year:ID, 38; ID, 17

Fixed effects:
          Estimate Std. Error t value Pr(>|z|)    
(Intercept)    4.07652    0.10395  39.217  < 2e-16 ***
park1          0.35882    0.10381   3.456 0.000548 ***
season1       -0.08571    0.13749  -0.623 0.533024    
season2        0.06915    0.12411   0.557 0.577380    
season3        0.04081    0.12784   0.319 0.749554    
season4        0.51415    0.20602   2.496 0.012574 *  
season5        0.01990    0.13948   0.143 0.886525    
season6       -0.55669    0.22691  -2.453 0.014152 *  
season7       -0.32262    0.18747  -1.721 0.085256 .  
park1:season1 -0.22670    0.13761  -1.647 0.099482 .  
park1:season2 -0.18563    0.12389  -1.498 0.134042    
park1:season3 -0.25638    0.12789  -2.005 0.045005 *  
park1:season4  0.92146    0.20639   4.465 8.02e-06 ***
park1:season5 -0.17337    0.13925  -1.245 0.213114    
park1:season6 -0.10128    0.22720  -0.446 0.655756    
park1:season7  0.10990    0.18755   0.586 0.557901    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
$\endgroup$
  • $\begingroup$ how did you estimate your model? If I use ` family = gaussian("log")` within glmer, I get the error, "response must be numeric". Maybe you can add a line with the command to your question? $\endgroup$ – Qaswed Nov 28 '18 at 11:06
0
$\begingroup$

Season estimates are set in referance to season 8..same with park levels are set in reference to park 2...i.e. the estimate for these are 0. This is a result of you using categorical effects. So to get a prediction for park 2 in season 8, it would just be the intercept. Interpreting the interaction term park1 by season 3 is more ambiguous with categorical effects. I would suggest doing tukey pairwise tests of park by season.

$\endgroup$
  • $\begingroup$ Thanks, that clears things up a bit. So, for an estimate of park 1 in season 8 as well, a pairwise test would help? $\endgroup$ – StephD Jun 25 '18 at 14:14
  • $\begingroup$ To compare to park 2 season 8 or to compare different seasons in park 1 $\endgroup$ – OliverFishCode Jun 25 '18 at 14:16
  • $\begingroup$ Yes, just realized previous comment may be unclear $\endgroup$ – OliverFishCode Jun 25 '18 at 16:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.