Hello everybody! I?m working with a dataset from twelve fertilizer trials, where the technical fertilizer product and application method, but not the intensity of fertilization, is varied. (I?m using R1.7.1 and W2000.) The call: ejna1t4b.lme <- lme( Yield ~ TrCode, data = ejna1t4, + random = ~ 1 | Trial/Block) works as far as I can understand well, the Block structure of the trials is used efficiently and everything looks nice according to plots of the object. Now I want to evaluate the influence of observations from the different experimental places (for example soil analyses or rainfall) - Could I do that without skipping the Trial/Block structure, or do I have to start from scratch again? The observed values will naturally only have one level for each Trial, so the term Trial/Block will host the effects of all observed (and unobserved) phenomena in each trial. Now I want to know where the effects come from. I?ve been looking for a text on this, both in MASS and Pinheiro & Bates, without finding any. Any hints of where to look? Thanks /CG CG Pettersson, MSci, PhD Stud. Swedish University of Agricultural Sciences Dep. of Ecology and Crop Production. Box 7043 SE-750 07 Uppsala
Might "update" help? spencer graves CG Pettersson wrote:>Hello everybody! >I?m working with a dataset from twelve fertilizer trials, where the >technical fertilizer product and application method, but not the >intensity of fertilization, is varied. (I?m using R1.7.1 and W2000.) > >The call: > >ejna1t4b.lme <- lme( Yield ~ TrCode, data = ejna1t4, >+ random = ~ 1 | Trial/Block) > >works as far as I can understand well, the Block structure of the >trials is used efficiently and everything looks nice according to >plots of the object. > >Now I want to evaluate the influence of observations from the >different experimental places (for example soil analyses or rainfall) >- Could I do that without skipping the Trial/Block structure, or do I >have to start from scratch again? The observed values will naturally >only have one level for each Trial, so the term Trial/Block will host >the effects of all observed (and unobserved) phenomena in each trial. >Now I want to know where the effects come from. > >I?ve been looking for a text on this, both in MASS and Pinheiro & >Bates, without finding any. Any hints of where to look? > >Thanks >/CG > >CG Pettersson, MSci, PhD Stud. >Swedish University of Agricultural Sciences >Dep. of Ecology and Crop Production. Box 7043 >SE-750 07 Uppsala > >______________________________________________ >R-help at stat.math.ethz.ch mailing list >https://www.stat.math.ethz.ch/mailman/listinfo/r-help > >
Oh, the question might have been more precisely formulated I guess... "update" surely helps a lot in the practical work at the computer. But that is not my problem. The problem is where and how to introduce the variables in the command. I?ve tried things like: random = ~ P.AL | Trial/Block as the random call, looking for some sort of direct effect from the soil analysis P.AL. That results in a worse model than without the term, possibly becouse the effect of the soil already is, implicitlely, inside the term Trial. What I am looking for is a way of evaluating several factors that have just one observation for each Trial, still having the the nice Trial/Block structure present in the model. Or am I just stupid? /CG -------------------> Might "update" help? spencer graves > > CG Pettersson wrote: > > >Hello everybody! > >I?m working with a dataset from twelve fertilizer trials, where the > >technical fertilizer product and application method, but not the > >intensity of fertilization, is varied. (I?m using R1.7.1 andW2000.)> > > >The call: > > > >ejna1t4b.lme <- lme( Yield ~ TrCode, data = ejna1t4, > >+ random = ~ 1 | Trial/Block) > > > >works as far as I can understand well, the Block structure of the > >trials is used efficiently and everything looks nice according to > >plots of the object. > > > >Now I want to evaluate the influence of observations from the > >different experimental places (for example soil analyses orrainfall)> >- Could I do that without skipping the Trial/Block structure, or doI> >have to start from scratch again? The observed values willnaturally> >only have one level for each Trial, so the term Trial/Block willhost> >the effects of all observed (and unobserved) phenomena in eachtrial.> >Now I want to know where the effects come from. > > > >I?ve been looking for a text on this, both in MASS and Pinheiro & > >Bates, without finding any. Any hints of where to look? > > > >Thanks > >/CG > > > >CG Pettersson, MSci, PhD Stud. > >Swedish University of Agricultural Sciences > >Dep. of Ecology and Crop Production. Box 7043 > >SE-750 07 Uppsala > > > >______________________________________________ > >R-help at stat.math.ethz.ch mailing list > >https://www.stat.math.ethz.ch/mailman/listinfo/r-help > > > > > >CG Pettersson, MSci, PhD Stud. Swedish University of Agricultural Sciences Dep. of Ecology and Crop Production. Box 7043 SE-750 07 Uppsala