Interesting topic! I had thought the details in the novel "The Story of Edgar Sawtelle" were a bit exaggerated, but apparently there's a lot more to it. I think Pedigrees is a fine name.
On Wednesday, December 17, 2014 2:57:14 AM UTC+7, Douglas Bates wrote: > > On Tuesday, December 16, 2014 1:32:19 PM UTC-6, Taylor Maxwell wrote: >> >> Pedigrees sounds like a good name. Will this package be able to generate >> and manipulate standard kinship matrices? This is not my specialty but >> familial data is pretty common and many standard association methods for >> "unrelated" individuals have analogous method that need to take into about >> the relationships among individuals that come with familial data. In >> particularly the the kinship matrix is often a large sparse matrix that is >> used as random effects in mixed linear models. The kinship package in R is >> very useful for the construction and use of these matrices. With some >> colleagues (who know the statistical theory better than I) we have >> developed a methodology to look at simultaneous mean and variance effects >> in unrelated individuals. We recently adapted it to use family data. We >> were able to use the kinship package in R to construct the kinship matrix >> (a sparse matrix) and applied our methodology with the help of their >> "gchol" function to implement a generalized Cholesky decomposition and used >> it along with BFGS to optimize the likelihood function. I have implemented >> the "unrelated" method in Julia and would love to be able to implement this >> family based method in Julia as well. >> > > Ana and I are familiar with using pedigree information in mixed-effects > models; we wrote the R package pedigreemm and I am one of the authors of > the lme4 package for R. The purpose of creating tools for pedigrees in > Julia is to use Pedigrees and MixedModels to fit such models using Julia. > > I should look again at the kinship package for R. (I get twitchy every > time I start to read it because Terry chose to reimplement sparse matrix > methods from scratch instead of using the recommended Matrix package for R > that Martin MÓ“chler and I wrote). I have the suspicion that the ideas > are the same, it's just the names that are different. > > On Tuesday, December 16, 2014 10:03:06 AM UTC-7, Douglas Bates wrote: >>> >>> Is anyone working on a Julia package or packages to work with pedigrees, >>> in the sense used in animal breeding? Ana Vazquez and I are developing >>> such a package which we plan to call Pedigrees. Its principle type will be >>> Pedigree, which is constructed from two Integer vectors representing the >>> 'sire' and 'dam' of each animal. These names reflect the animal breeding >>> orientation of the package. >>> >>> Capabilities include sorting a pedigree so that parents occur before >>> offspring and deriving the additive relationship matrix and other >>> quantities related to it, such as inbreeding coefficients. >>> >>> Is using the name Pedigrees going to cause conflicts? >>> >>
