Message-ID: <loom.20150427T153122-447@post.gmane.org>
Date: 2015-04-27T13:34:16Z
From: Ben Bolker
Subject: Results from GLMM, error bars for predictions
Quentin Schorpp <quentin.schorpp at ...> writes:
>
> Dear Thierry,
>
> There's one thing that i don't understand.
> Why do you plot the vertical line in
> ggplot(pairwise.ci <http://pairwise.ci>, aes(y = lhs, x = exp(estimate),
> xmin = exp(lwr), xmax = exp(upr))) + geom_errorbarh() + geom_point() +
> geom_vline(xintercept = 1)
>
> and how can i transfer this to other models?
>
> kind regards,
> Quentin
[Sorry to remove context, gmane doesn't like it ...]
Since the model parameters are back-transformed (i.e.
exponentiated), the null value of the parameters (no effect)
is exp(0)=1; the horizontal line is showing the null value.
Thus for example if the 95% confidence intervals cross the
dotted line, the corresponding effect is not statistically
significant at p=0.05.