[R] degrees of freedom for contrast

Qian Liu littleduck24 at gmail.com
Thu Jun 7 22:12:51 CEST 2012


Hi, I need some help to figure out the df I should use in t test for
my contrast.
I have 5 treatments and 5 phenotypes, I would like to compute the
difference of treatment means for each phenotype and do t test, such
as treatment1 vs treatment2 on phenotype1
How should I calculate the pooled degrees of freedom for the t tests
of all the contrasts? Thank you very much.

Qian



>mylong.lme <- lme(dscore~Trt.Pheno-1, data=mylong, random=~1 | ID, method="ML")
> summary(mylong.lme)
Linear mixed-effects model fit by maximum likelihood
 Data: mylong
       AIC      BIC    logLik
  14789.14 14949.83 -7367.571

Random effects:
 Formula: ~1 | ID
        (Intercept) Residual
StdDev:     1.40765 3.039555

Fixed effects: dscore~ Trt.Pheno - 1
                                       Value Std.Error   DF    t-value p-value
TrtPheno1_1 :    -2.516975 0.2788703 2412  -9.025613  0.0000
Trt.Pheno2_1 :  -1.172767 0.3781179 2412  -3.101590  0.0019
Trt.Pheno3_1 :    -0.810177 0.2869447 2412  -2.823459  0.0048
Trt.Pheno4_1 :   -1.518063 0.2791157 2412  -5.438830  0.0000
Trt.Pheno5_1 :   -0.367947 0.3564081 2412  -1.032377  0.3020
..............
.............
coef <- fixed.effects(mylong.lme)
covmat <- mylong.lme$varFix
  > c # my contrast matrix
  >mycontr.est <- c %*% coef
  >mycontr.var <- c %*% covmat %*% t(c)
  >t2 <- t(mycontr.est) %*% solve(mycontr.var) %*% mycontr.est
  >P<- 2*pt(sqrt(abs(t2)), df=????)
> version
               _
platform       i386-pc-mingw32
arch           i386
os             mingw32
system         i386, mingw32
status
major          2
minor          14.0
year           2011
month          10
day            31
svn rev        57496
language       R
version.string R version 2.14.0 (2011-10-31)



More information about the R-help mailing list