2
$\begingroup$

How are degrees of freedom calculated for a linear mixed effects model? I was recently running a linear mixed effects model that contained 1 independent variable with 3 levels (1,2, and 3). The random effects included: individual animal ID and year as I expected these variables to account for some of the variability in the modelling procedure and wanted to make sure to account for it. However, to obtain the P-value, I ran a null model containing only the intercept and random effects and compared it via the anova() command to the full model. The degrees of freedom just seem a little odd given that the random effect for individual animal ID contains 171 individuals across 3 years. Any help would be much appreciated!

fit.lmer.model <- lmer(MCP ~ Season +(1|ID)+(1|Year), data=mydataframe,REML=FALSE)
summary(fit.lmer.model) 

fit.lmer.null <- lmer(MCP ~ 1 + (1|ID)+(1|Year), data=mydataframe,REML=FALSE)
summary(fit.lmer.null) 

anova(fit.lmer.null,fit.lmer.model)

> anova(fit.lmer.null,fit.lmer.model)
Data: mydataframe
Models:
fit.lmer.null: MCP ~ 1 + (1 | ID) + (1 | Year)
fit.lmer.model: MCP ~ Season + (1 | ID) + (1 | Year)
           Df    AIC    BIC  logLik deviance Chisq Chi Df Pr(>Chisq)  
fit.lmer.null   4 2240.0 2252.6 -1116.0   2232.0                          
fit.lmer.model  5 2238.2 2253.9 -1114.1   2228.2 3.759      1    0.05252 .
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
$\endgroup$
1
$\begingroup$

The degrees of freedom shown are the number of estimated parameters in the model. In the fit.lmer.null model you have a parameter for the intercept, a variance parameter for each random effect term and a variance parameter for the observations given the random effects (called "Residual" if you run summary(fit.lmer.null)). That is a total of four parameters. The principle is the same for the other model, you just have one more parameter from "Season".

$\endgroup$
1
  • $\begingroup$ Thank you very much for providing some explanation on the degrees of freedom calculation in a linear mixed model framework! $\endgroup$
    – Buck2079
    Apr 2 '15 at 16:09

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.