[R-sig-ME] Repeated measures ANOVA using lmer, nlme, and geepack

[Paul Miller]

Hi John,

Thanks for your reply. Interestingly, the coefficients do not match (see below). 

Hopefully, someone won't mind explaining why.

If not, I may figure it out at some point. I've been reading Data Analysis using Regression and Multilevel/Hierarchical Models by Gelman and Hill. I'm getting aquainted with my new cartoon character friends Elmer (lmer) and Bugs, and at some point it may become clear what is going on. At the moment though, the whole thing seems a bit Looney Tunes.

Paul


> model.un <- lme(ADASCOG ~ TREAT * MONTH, random = ~ 1 | PAT,  
+  correlation=corSymm(form = ~ 1 | PAT), 
+  data=unialz, na.action=na.omit)
> summary(model.un)

Fixed effects: ADASCOG ~ TREAT * MONTH 
                         Value Std.Error  DF   t-value p-value
(Intercept)           41.35112 1.6602811 359 24.906097  0.0000
TREATHigh             -5.62566 2.4823037  77 -2.266307  0.0262
TREATLow              -3.52850 2.4265281  77 -1.454134  0.1500
MONTHADAS02           -9.41879 1.1007344 359 -8.556821  0.0000
MONTHADAS04           -6.48828 1.1776015 359 -5.509738  0.0000
MONTHADAS06           -4.69629 1.1362679 359 -4.133084  0.0000
MONTHADAS08           -2.19226 1.0765468 359 -2.036378  0.0424
MONTHADAS10           -1.50837 0.7758601 359 -1.944132  0.0527
TREATHigh:MONTHADAS02  5.10358 1.6322908 359  3.126637  0.0019
TREATLow:MONTHADAS02   2.83802 1.5963900 359  1.777776  0.0763
TREATHigh:MONTHADAS04  4.64845 1.7510112 359  2.654725  0.0083
TREATLow:MONTHADAS04   2.46563 1.7243495 359  1.429892  0.1536
TREATHigh:MONTHADAS06  4.59431 1.7116064 359  2.684209  0.0076
TREATLow:MONTHADAS06   2.91622 1.6535434 359  1.763618  0.0786
TREATHigh:MONTHADAS08  1.47826 1.5829411 359  0.933872  0.3510
TREATLow:MONTHADAS08  -0.86712 1.5609770 359 -0.555497  0.5789
TREATHigh:MONTHADAS10  0.83351 1.1515076 359  0.723841  0.4696
TREATLow:MONTHADAS10   0.61007 1.1416625 359  0.534368  0.5934


PROC MIXED DATA = UNIALZ;
    CLASS TREAT MONTH PAT;
    MODEL ADASCOG = TREAT MONTH TREAT*MONTH / S;
    REPEATED / TYPE=UN SUBJECT=PAT RCORR;
RUN;

                     Solution for Fixed Effects

                                        Standard
Effect       TREAT  MONTH  Estimate     Error    DF  t Value  Pr > |t|

Intercept                   41.5542    1.7842    77    23.29    <.0001
TREAT        H              -5.8322    2.6666    77    -2.19    0.0318
TREAT        L              -3.6540    2.6086    77    -1.40    0.1653
TREAT        P                    0         .     .      .       .
MONTH                2      -9.4615    1.1800    77    -8.02    <.0001
MONTH                4      -6.7188    1.2828    77    -5.24    <.0001
MONTH                6      -4.8139    1.2488    77    -3.85    0.0002
MONTH                8      -2.3557    1.0647    77    -2.21    0.0299
MONTH               10      -1.7558    0.7352    77    -2.39    0.0194
MONTH               12            0         .     .      .       .
TREAT*MONTH  H       2       5.3228    1.7497    77     3.04    0.0032
TREAT*MONTH  H       4       5.1414    1.9054    77     2.70    0.0086
TREAT*MONTH  H       6       4.4064    1.8677    77     2.36    0.0209
TREAT*MONTH  H       8       1.5921    1.5706    77     1.01    0.3139
TREAT*MONTH  H      10       1.2032    1.0963    77     1.10    0.2758
TREAT*MONTH  H      12            0         .     .      .       .
TREAT*MONTH  L       2       2.7943    1.7174    77     1.63    0.1078
TREAT*MONTH  L       4       2.5602    1.8728    77     1.37    0.1756
TREAT*MONTH  L       6       3.0176    1.8194    77     1.66    0.1013
TREAT*MONTH  L       8      -0.7675    1.5415    77    -0.50    0.6200
TREAT*MONTH  L      10       0.8356    1.0704    77     0.78    0.4374
TREAT*MONTH  L      12            0         .     .      .       .
TREAT*MONTH  P       2            0         .     .      .       .
TREAT*MONTH  P       4            0         .     .      .       .
TREAT*MONTH  P       6            0         .     .      .       .
TREAT*MONTH  P       8            0         .     .      .       .
TREAT*MONTH  P      10            0         .     .      .       .
TREAT*MONTH  P      12            0         .     .      .       .