# Changing reference level on mixed model changes pMCMC value

I'm using an lmer on a set of continuous data. This post helped me a great deal when interpreting results, as I have two conditions and three levels as well, but I've run into a frustrating problem.

My data is from French speakers in two conditions: 1) with an English interviewer and 2) with a French interviewer. They are either intermediate, advanced, or native proficiency.

When the reference level is "Eng/Intermediate," I get different pMCMC values than when the reflevel is "Fr/Intermediate," between the same groups. I starred what should be the p-values for Intermediate vs. Advanced in the French condition as an example. The first set of R output is with group = eng and proficiency = intermediate as the reflevels, and the second is with group = fr and proficiency = intermediate as the reflevels.

                           Estimate  MCMCmean HPD95lower  HPD95upper pMCMC   Pr(>|t|)
(Intercept)                  2254.16  2253.65    2148.11    2362.66 0.0001   0.0000
groupfr                       129.93   130.18      81.42     179.85 0.0001   0.0000
proficiencyadvanced          -195.87  -195.60    -331.20     -58.28 0.0082   0.0080
proficiencynative             -80.76   -80.36    -235.26      70.25 0.2780   0.3241
groupfr:proficiencyadvanced    75.94    75.66      11.75     138.42 **0.0184**   0..0195
groupfr:proficiencynative     -58.95   -59.60    -130.00      12.38 0.1006   0.1017

Estimate MCMCmean HPD95lower HPD95upper  pMCMC Pr(>|t|)
(Intercept)                   2384.10  2384.86    2280.74    2495.84 0.0001   0.0000
groupeng                      -129.93  -130.61    -180.48     -80.20 0.0001   0.0000
proficiencyadvanced           -119.92  -120.64    -257.98      17.45 **0.0812**   0.1051
proficiencynative             -139.72  -140.22    -293.25      12.33 0.0708   0.0885
groupeng:proficiencyadvanced   -75.94   -74.91    -137.69     -10.48 0.0204   0.0195
groupeng:proficiencynative      58.95    59.77     -11.66     129.56 0.0974   0.1017


Am I misinterpreting the output of pvals.fnc? Thanks in advance for any help.

-