Cranial measurements and individual information for mice from the long-term
artificial selection experiment on body weight analyzed by Penna et al.
(2017). Selection for body weight at 49 days of age was carried out
downwards in the s and h lines and upwards in the sp and
hp lines, alongside a randomly mated control line t; the
s/sp lines were inbred by limiting population size and the
h/hp lines by full-sib mating in the first generations.
Format
A data frame with 329 rows and 46 columns. Row names are the
individual IDs (as.character(ID)).
.idline label combining the selection direction and the line, e.g.
"control.t";IDindividual catalogue number;
SEXsex,
"F"or"M";GERgeneration of the long-term experiment;
LINexperimental line:
t(control),sandh(downward selection),spandhp(upward selection);MADREcatalogue number of the dam;
PADREcatalogue number of the sire;
AGEage at sacrifice, in days;
P49body weight at 49 days of age;
linethe line as a factor with levels
t,h'}, \code{s',hands;selectiondirection of selection:
"control","downwards"or"upwards";IS_PM,IS_NSL,IS_PNS,PM_ZS,PM_ZI,PM_MT,NSL_NA,NSL_ZS,NSL_ZI,NA_BR,NA_PNS,BR_PT,BR_APET,PT_APET,PT_BA,PT_EAM,PT_ZYGO,PT_TSP,ZS_ZI,ZI_MT,ZI_ZYGO,ZI_TSP,MT_PNS,PNS_APET,APET_BA,APET_TS,BA_EAM,EAM_ZYGO,ZYGO_TSP,LD_AS,BR_LD,OPI_LD,PT_AS,JP_AS,BA_OPI35 cranial Euclidean distances, named after the landmarks they span. They are on the measured scale;
log()is the usual transformation before fitting.
Source
Penna et al. (2017), archived at Dryad under CC0 1.0 (doi:10.5061/dryad.5gr8r ) and distributed by the MIT-licensed companion package https://github.com/diogro/ratones.
Details
Every individual carries a full set of 35 cranial Euclidean distances, measured as the mean of two replicas, plus the catalogue numbers of its parents. The distances are moderately integrated: the first principal component of their correlation matrix explains about 40% of the variance and the mean pairwise correlation is 0.35, so the traits carry partly independent information and a handful of them is enough for a well-identified model.
Weak identification
This is a long-term artificial-selection experiment that happens to carry a
pedigree, not a design built to estimate a genetic covariance matrix. The
animal model fits, but 329 individuals from 58 sires identify a 35-trait
\(\mathbf{G}\) only weakly: the split between the genetic and the
environmental covariance is left largely to the priors, so a heritability
read from lmm_animal on these traits depends on the prior as
much as on the data, and can move by a factor of eight between defensible
choices. Use it as the small worked example it is, and use
opossum, which carries marker genotypes and a design that does
identify \(\mathbf{G}\), for anything about genetic covariance.
References
Penna, A., Melo, D., Bernardi, S., Oyarzabal, M. I., & Marroig, G. (2017). The evolution of phenotypic integration: How directional selection reshapes covariation in mice. Evolution, 71(10), 2370-2380. doi:10.1111/evo.13304
