[R-lang] Plotting CI from Mixed Effects Model

[Corey McMillan]

Hi-

I am reporting results from a mixed effects model in which Place (-,+) and 
Voice (-,+) are fixed effect factors.  I would like to plot the coefficient 
estimates along with their 95% CIs to visualise my results.  The model and 
MCMC estimations look like this:

artic.lmer <- lmer(nonorm ~ relevel(artic5$place, "place-") * 
relevel(voice, "voice-") + (1 | sub) + (1 | item), data=artic5)

pvals.fnc(artic.lmer)

$fixed
															Estimate
(Intercept)													1.7156
relevel(artic5$place, "place-")place+							1.4482
relevel(voice, "voice-")voice+ 									0.9149
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+  -1.0889
															MCMCmean
(Intercept)													1.7169
relevel(artic5$place, "place-")place+ 							1.4504
relevel(voice, "voice-")voice+									0.9147
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+  -1.0918
															HPD95lower
(Intercept)													1.0303
relevel(artic5$place, "place-")place+							0.9469
relevel(voice, "voice-")voice+									0.3433
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+	-1.6631
															HPD95upper
(Intercept)													2.4780
relevel(artic5$place, "place-")place+							1.9432
relevel(voice, "voice-")voice+									1.4374
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+	-0.4564
															pMCMC
(Intercept)													0.0012
relevel(artic5$place, "place-")place+							0.0001
relevel(voice, "voice-")voice+									0.0012
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+	0.0006
															Pr(>|t|)
(Intercept)													0.0000
relevel(artic5$place, "place-")place+							0.0000
relevel(voice, "voice-")voice+									0.0009
relevel(artic5$place, "place-")place+:relevel(voice, "voice-")voice+	0.0004

My understanding is that the value for each cell in the 2 X 2 design can be 
calculated in the following way:

place-,voice-   = 1.7156
place-,voice+  = 1.7516 + 0.9149 = 2.67
place+, voice- = 1.7516 + 1.4482 = 3.20
place+,voice+ = 1.7516 + 0.9149 + 1.4482 - 1.0889 =  3.03

Now if I want to report the 95% CIs around those values can I use the same 
procedure?  So the place-,voice+ lower CI would be equivalent to 0.9469 + 
1.7156 and the higher CI equivalent to 1.943 + 1.7156, and so on...

Any clarifications on this would be greatly appreciated.

Cheers,
Corey McMillan



-- 
Corey McMillan
University of Edinburgh
School of Philosophy, Psychology, and Language Sciences

email: Corey.McMillan at ed.ac.uk
web: http://homepages.ed.ac.uk/s0340151