1

I obtained data of typical time point in two conditions in this shape:

weight   condition   time
0.1307857    Transf    1
0.1926429    Transf    2
0.2734286    Transf    3
0.4403571    Transf    4
0.6037143    Transf    5
0.9036429    Transf    6
1.5454286    Transf    7
0.1370714       Unt    1
0.2005000       Unt    2
0.2973571       Unt    3
0.4592143       Unt    4
0.8336429       Unt    5
1.3099286       Unt    6
2.1470000       Unt    7

I am using the package nlme in R to test the difference between the two conditions in time

fm1 <- lme(weigth~condition,random=~1|time,data=data_new)
anova.lme(fm1, adjustSigma = F)

I got this:

numDF denDF  F-value p-value
(Intercept)     1     6 8.421439  0.0273
condition       1     6 4.208786  0.0861

Questions:

  1. Am I doing things correctly with this procedure to test the differences between the two conditions in time?

  2. If yes, which is my p-value, the intercept one or the condition?

  3. If this is not the proper test, what are the alternatives?

Randel
  • 6,199
  • 4
  • 39
  • 65
ToniG
  • 125
  • 1
  • 1
  • 7
  • 1
    The p-value corresponding to condition is the one that relates to differences in condition. I notice the variable 'weight' is actually spelled 'weigth' in your code. From bitter experience, this sort of misspelling in code can lead to a lot of misery, even if you're consistent about it. – Glen_b Jun 07 '13 at 12:57

1 Answers1

1

Sounds to me like your goal is to model the trajectories and check for an effect of both time and condition. If that is the case (and I'm assuming your dataset is larger with multiple subjects, not just aggregated by time and condition), you would want something like:

lme(weigth~time*condition, random=~1|subject, data=data_new)

Typically you want to build a mixed-effects model from the bottom up. See page 66 of this for a good step-by-step for nlme for growth curves.

dmartin
  • 3,010
  • 3
  • 22
  • 27