R help - Oct 2003 - Evaluating outer, numeric, variables in an lme object.

Hello!
I?m working with a dateset from twelve fertilizer experiments (Trial) with 
a block structure of three replicats (Block). The treatment levels consist 
of application method and product but only one intensity. The factor TrCode 
could for example be BC(CAN) for broadcast calcium ammonium nitrate. I?ve 
used the following call to fit a lme-object to the data:

ejna1t4b.lme <- lme( Yield ~ TrCode, data = ejna1t4,
+                       random = ~ 1 | Trial/Block)

This seems, for me, to work quite well. Or should I code differently?

What I want to do now, is to evaluate the impact from a load of soil and 
weather data where I only have one reading from each Trial/variable. These 
variables are all numeric.

Is this possible in the lme context?
It is naturally possible to calculate treatment means of all Trials and run 
a regression analysis on the new dataset, but I have a feeling there are 
smarter ways of doing this. Or?

Is there anything on this type of problem in MASS or Pinheiro & Bates? So 
far, if I have found the proper texts, I have not understood it... ;-)

Thanks
/CG



CG Pettersson, MSc, PhD-stud
Department of Ecology and Crop Production Science, SLU
P.O. Box 7043
S-750 07 Uppsala, Sweden
Phone: +46 (0)18 67 12 24; Fax: +46 (0)18 67 29 06
        +46 (0)70 330 66 85