1

I've trying to understand the meaning of the significant and negative interaction between a categorical and a continuous variable in the model I attach here. This model was the best model obtained after AIC model selection (weight=1). The response variable MPO refers to the use of the microhabitat fern by individuals, that is: MPO=0 when fern was not used and MPO=1 when it was used. The explanatory variables are treatment (categorical: RC and NT) and SVL (size of individuals, continuous).

 Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
 Family: binomial  ( logit )
Formula: MPO ~ TREAT * SVL + (1 | GRID/ID)
   Data: arbino5
Control: glmerControl(optimizer = "optimx", optCtrl = list(method = "nlminb"))

 AIC      BIC   logLik deviance df.resid 
1764.4   1804.3   -876.2   1752.4     5682 

Scaled residuals: 
    Min      1Q  Median      3Q     Max 
-1.7812 -0.0469 -0.0304 -0.0249  6.8455 

Random effects:
 Groups  Name        Variance Std.Dev.
 ID:GRID (Intercept) 15.56    3.945   
 GRID    (Intercept)  0.00    0.000   
Number of obs: 5688, groups:  ID:GRID, 2219; GRID, 4

Fixed effects:
             Estimate Std. Error z value Pr(>|z|)    
(Intercept) -11.68453    1.15979 -10.075  < 2e-16 ***
TREATRC       3.09034    1.06935   2.890  0.00385 ** 
SVL           0.26491    0.03368   7.865 3.68e-15 ***
TREATRC:SVL  -0.18267    0.04172  -4.378 1.20e-05 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Correlation of Fixed Effects:
            (Intr) TREATRC SVL   
TREATRC     -0.775               
SVL         -0.878  0.811        
TREATRC:SVL  0.747 -0.957  -0.825

I cant fully understand what this interaction means, specially considering that the main effects have positive estimates. I tried using 

interaction.plot(arbino5$SVL, arbino5$TREAT, arbino5$MPO, type ="p")

to graph this interaction but it wasn't helpful as SVL has tons of points.

Could anyone help me with this? Thanks so much

Patty
  • 11
  • 1
  • Which interaction? The correlation of fixed effects, perhaps? – Carl Sep 20 '16 at 01:40
  • Maybe Im not using the proper notation, Im fairly new to R, but I mean the `TREATRC:SVL` interaction – Patty Sep 20 '16 at 21:41
  • For correlation of fixed effects see link http://stats.stackexchange.com/questions/57240/how-do-i-interpret-the-correlations-of-fixed-effects-in-my-glmer-output – Carl Sep 20 '16 at 22:17
  • Interpretation of interactions is the same in linear and logistic regression – kjetil b halvorsen Sep 07 '17 at 12:01

0 Answers0