https://coffeehouse.dataone.org/2014/04/09/abandon-all-hope-ye-who-enter-dates-in-excel/

]]>Thank you very much by the answer.

Excuse the ignorance on the subject, but I am now initiating studies on Bayesian inference. I’d like to ask a question:

I want to define a priori information, with normal distribution for the data, in MCMCglmm package. So, what are the values that should be assigned to “V” ans “n” in argument “list (V = …. n …. =)” ??

Already thank you very much!!

]]>Luis (with s) here. While in this post I discuss the issue of overlapping matrices in the context of a diallel cross, I also point out that “With the advent of animal model BLUP, was possible to fit something like y = mu + blocks + individual (using a pedigree) + cross + error”. Fitting separate mums and dads predates the use of animal models, it is old-fashioned (we’re talking 30-40 years old) and much more complicated. I have code for the analysis of diallels in asreml here. In the case of MCMCglmm there are a few differences, starting with the pedigree terms HAS to be called animal. You could set up something like:

prior = list(R = list(V = 0.007, n = 0),

G = list(G1 = list(V = 0.002, n = 0), G2 = list(V = 0.001, n = 0), G3 = list(V = 0.001, n = 0))))

# I use high thinning to avoid autocorrelation with animal model

nvel.bayes < - MCMCglmm(nvel ~ 1,

random = ~ animal + Block + Family,

family = ‘gaussian’,

data = ncs,

pedigree = ped,

prior = prior,

verbose = FALSE,

pr = TRUE,

burnin = 10000,

nitt = 200000,

thin = 200)

ped is a data frame with the pedigree, with columns animal, Mum, Dad. This should fit the additive and family effects in a randomized complete block design. Reciprocal effects are home work.

]]>I am conducting a diallel analysis, and would to employ the mixed models based on Bayesian approach with MCMCglmm package.

But I’m struggling to define the model because of the sources of variation dad and mum, which must be defined in the model together so that a single variance component is obtained for both (dad and mum), as you defined in the your model, observed in the model m1 detailed below.

# Fitting the model with ASReml

library(asreml)

m1 <- asreml(yield ~ 1,

random = ~ block + dad + and(mum) + mate,

data = trial)

summary(m1)

# gamma component std.error z.ratio

#block!block.var 1.299110e-02 3.861892e-01 1.588423e+00 0.2431274

#dad!dad.var 2.101417e-01 6.246930e+00 5.120745e+00 1.2199259

#mate!mate.var 4.589938e-07 1.364461e-05 2.340032e-06 5.8309519

#R!variance 1.000000e+00 2.972722e+01 5.098177e+00 5.8309519

But I did not get information about this possibility for MCMCglmm package in R. It would be possible to set the mcmc model for MCMCglmm also detailed below, in the same way that is set to m1 model, the ASReml?

# Fitting the model with MCMCglmm

library(MCMCglmm)

prior.mcmc <- list(R = list(V = 1e-16, nu = -2),

G = list(G1 = list(V = 1e-16, nu = -2),

G2 = list(V = 1e-16, nu = -2),

G3 = list(V = 1e-16, nu = -2)))

mcmc = MCMCglmm(yield ~ 1,

random = ~ block + dad + mum + mate,

family = 'gaussian',

data = trial,

prior = prior.mcmc,

verbose = FALSE,

pr = TRUE,

burnin = 10000,

nitt = 20000,

thin = 10)

summary(mcmc)

Already very grateful for your attention!

]]>One has to understand though, that there are some cases where classical breeding is painfully slow or difficult. For example, improving two negatively correlated traits could be achieved more easily by using modification or, at least, using molecular genetics to track the genes of interest. Anyway, there are plenty of problems where classical breeding will do very well.

]]>