[R-sig-ME] Discrepancy between STATISTICA results and lmer and lme
Mario Garrido
g@@d|o @end|ng |rom po@t@bgu@@c@||
Tue Sep 3 13:42:24 CEST 2019
Morning,
I am running a linear mixed model withrepeated measures using lmer and I
have found that results are the same whe I use lme or aov with repeated
measure. However, are slightly different for STATISTICA software. I guess
is something related with the estimation method or any kind of correction
but I would like to now the reason and I am not able to find it.
Anybody knows or have any recommended reading at this respect?
Thanks in advance
With lmer from package lme4
lmer5 <- lmer(Mean ~ Trtmnt*sp*Period + (1|ExpID), data =
Data[Period%in%c("Before","early peak","late peak"),])
anova(lmer5)
Analysis of Variance Table
Df Sum Sq Mean Sq F value
Trtmnt 2 2.80 1.398 0.6762
sp 2 592.64 296.319 143.3659
Period 2 6.28 3.141 1.5195
Trtmnt:sp 4 12.70 3.176 1.5366
Trtmnt:Period 4 41.99 10.499 5.0795
sp:Period 4 4.61 1.151 0.5571
Trtmnt:sp:Period 8 13.83 1.729 0.8364
With lme from package nlme
lme5 <- lme( Mean ~ Trtmnt*sp*Period, data =
Data[Period%in%c("Before","early peak","late peak"),], random = ~
1|factor(ExpID))
anova(lme5)
numDF denDF F-value p-value
(Intercept) 1 122 5739.446 <.0001
Trtmnt 2 61 0.676 0.5123
sp 2 61 143.366 <.0001
Period 2 122 1.519 0.2229
Trtmnt:sp 4 61 1.537 0.2029
Trtmnt:Period 4 122 5.079 0.0008
sp:Period 4 122 0.557 0.6942
Trtmnt:sp:Period 8 122 0.836 0.5724
With STATISTICA
Effect Repeated Measures Analysis of Variance (Lab
coinfection_averages_BM.sta)
Sigma-restricted parameterization
Effective hypothesis decomposition; Std. Error of Estimate: 2.4177
SS Degr. of MS F p
Intercept 395317,0 1 395317,0 5173,785 0,000000
Trtmnt 169,0 2 84,5 1,106 0,337542
sp 18404,8 2 9202,4 120,438 0,000000
Trtmnt*sp 469,6 4 117,4 1,537 0,202864
Error 4660,9 61 76,4
PERIOD 3,8 2 1,9 0,923 0,399914
PERIOD*Trtmnt 42,5 4 10,6 5,146 0,000726
PERIOD*sp 3,6 4 0,9 0,440 0,779813
PERIOD*Trtmnt*sp 13,8 8 1,7 0,836 0,572377
Error 252,2 122 2,1
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list