[R-sig-ME] Are BLUP from two different random factors independant ?

[LASSALVY Stephane]

Dear Thierry,
Yes you are right. I think it would be a way of doing it.

Best regards,
Stephane

De : Thierry Onkelinx <mailto:thierry.onkelinx using inbo.be> 
Envoyé : mardi 27 novembre 2018 09:59
À : LASSALVY Stephane <mailto:stephane.lassalvy using geves.fr>
Cc : r-sig-mixed-models <mailto:r-sig-mixed-models using r-project.org>
Objet : Re: [R-sig-ME] Are BLUP from two different random factors independant ?

Dear Stephane,

Why not simply fit a model without  [[variety x network]]. Then IMHO [[variety x year x network]] would be an estimate of sum you're interested in.

Best regards,

ir. Thierry Onkelinx
Statisticus / Statistician

Vlaamse Overheid / Government of Flanders
INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND FOREST
Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance 
mailto:thierry.onkelinx using inbo.be
Havenlaan 88 bus 73, 1000 Brussel
http://www.inbo.be
///////////////////////////////////////////////////////////////////////////////////////////
To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher
The plural of anecdote is not data. ~ Roger Brinner
The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey
///////////////////////////////////////////////////////////////////////////////////////////

https://www.inbo.be/


Op di 27 nov. 2018 om 09:53 schreef LASSALVY Stephane <mailto:stephane.lassalvy using geves.fr>:
Hello all,
I am encountering a question related to linear combination of the estimations of random effects with the Asreml-R package. I am considering the problem of getting the sum of two BLUP. I only want to make the sum of two random effects.

My team and I, we are modelling over the years and over networks field trials data. The number of years which we use for an analysis is usually 2 and the number of networks is 3.

The model which we use is :
Yield = intercept + year + network + year x network + year x network x trial + [[variety]] + [[variety x network]] + [[variety x year x network]] + [[variety x year x network x trial]] + [[error term ]]

Year, network, year x network are fixed effects. The terms between brackets [[ ]] are random terms so we can compute BLUP.

We get the estimation of random terms for single random terms with the summary.asreml command with the all = TRUE option.


  1.  Would you know a mean to compute the sum of two random terms estimations ? We are mostly interested in computing the sum of a BLUP "variety x network" and the related BLUPs for the "variety x year x network" term to get the estimation of the variety x network interaction for each year of trial. Computing the sum of the estimations is not so difficult, but I am not sure that we will be able to compute the standard error for such a sum. Therefore, if a function like predict is able to compute this sum, it would be a very good news for us.



  1.  In case I would have to do the calculation by myself, can I consider that, as  the random factors [[variety x network]]  and  [[variety x year x network]] are independent, for a given variety, BLUP(variety x network) and BLUP(variety x year x network) ?

Best regards,
Stephane

Stéphane LASSALVY
GEVES La Valette
711, rue Jean-François Breton
F-34090 Montpellier
Tél. +33 (0)4 67 04 35 81
Email : mailto:stephane.lassalvy using geves.fr<mailto:mailto:contact using geves.fr>


        [[alternative HTML version deleted]]

_______________________________________________
mailto:R-sig-mixed-models using r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models