qb.scanone {qtlbim} | R Documentation |
This method extracts iteration diagnostics and mainloci from the
qb
object and returns a data frame (of class qb.scanone
)
containing information on environmental variance, explained variance
components, non-epistatic variance components.
qb.scanone(qbObject, epistasis = TRUE, scan, type.scan, covar, adjust.covar, chr, sum.scan = "yes", min.iter = 1, aggregate = TRUE, smooth = 3, weight = c("sqrt","count","none"), split.chr, center.type = c("mode","mean","scan"), half = FALSE, verbose = FALSE, ...)
qbObject |
An object of class qb . |
epistasis |
If TRUE then information about
epistasis is included. |
scan |
Vector of diagnostics to scan (see below). |
type.scan |
Type of scan; default is "heritability" (see below). |
covar |
Covariate(s) to include; default is seq(nfixcov)
where nfixcov is taken from qb.data . Set to 0
to exclude any covariates. |
adjust.covar |
Adjustments to covariates. Default is
NA , which adjusts by covariate mean values. Values are
assumed to be in order of fixed covariates. |
chr |
Chromosomes to subset on if not NULL . |
sum.scan |
Sum over scan diagnostics if "yes" or "only";
only report sum if "only". |
min.iter |
Include only samples at loci if minimum number of
iterations is at least min.iter ; default is to include all
(min.iter = 1). |
aggregate |
Aggregate effects into main, epis, gbye if TRUE . |
half |
Cut epistatic effects in half if TRUE . |
smooth |
Degree of nearest neighbor smoothing to determine maxima. |
weight |
Weights to use for nearest neighbor
smoothing. sqrt is square root of count per locus. Used only
if smooth > 0. |
center.type |
Method to find QTL loci. See details. |
split.chr |
Split summary by multiple QTL per chromosome (see details). |
verbose |
Give verbose feedback if TRUE . |
... |
Additional arguments mostly ignored. |
The type.scan
specifies what type of scan is performed. Scan produces
marginal estimates of diagnostics at each potential loci across the
genome. That is, values are adjusted for other possible QTL simply by
taking the marginal average over MCMC samples. Choices of type.scan
are
"heritability", "LPD", "LR", "deviance", "detection", "variance",
"estimate", "cellmean", "count", "log10", "posterior", "logposterior"
(i.e. log10(posterior)), "BF", "2logBF" (i.e. 2*ln(BF)), and "nqtl"
(number of linked QTL). Default is "LPD".
Type "heritability" is actually R-squared at this point, not the theoretical
heritability.
Types "LPD", "LR" and "deviance" are all proportional to each other in
the usual sense; "LPD" is computed to agree with lod
from
scanone
if models were restricted to one QTL and missing
genotypes are imputed. Detection is the marginal posterior probability of
detectio of a QTL at a locus.
Types "variance" and "estimate" yield, respectively, the marginal variance
components and the marginal parameter estimates at each loci. Type
"cellmean" gives marginal estimates for A, H, B genotypes (these are
single character codes for AA, AB, BB, respectively). The remaining
count types provide diagnostics. Types "count" and "log10" report on
number of MCMC samples in raw or logged scale. Type
"posterior" ("logposterior") yields the marginal (log) posterior
probability. Type "BF" ("2logBF") gives the marginal
Bayes factor per loci; both are proportional to "count". Type "nqtl"
gives the average number of linked loci, which can be useful in sorting
out multiple linked loci.
The scan
specifies the model effects to include for all types
except the counts. Aggregated effects (default except for type
"cellmean") are "main", "epistasis" and "GxE" (genotype by
environment). Individual model effects can be requested
as "add", "dom", "aa", "ad", "da", "dd". In addition, GxE terms, if
present are included automatically if covar
is not 0. For type
"estimate", main effects for "add" and "dom" are adjusted for any
covariate GxE effects. The sum.scan
is used for all types but the
counts to get a summary across scan
effects.
The "mode"
and "mean"
centering rely on the mode and mean,
respectively, of the posterior; the "scan"
centering uses the
mode of the actual type used to create the qb.scanone
object. The "scan"
agrees with an scanone
summary method for "pos"
and "sum"
columns. However, the
mode for a "scan"
could be in a region of low posterior mass
and may not be reliable as such. Note that mean
can be
biased when there are linked loci. Only used in qb.scanone
summary.
Evidence for linked loci leads to multiple summary lines per
chromosome. Be default, a qbObject
has inferred chromosome splits
based on MCMC samples (see qb.split.chr
). This is
determined in a similar manner to
qb.multloci
. In particular,
the arguments cutoff
and nqtl
documented in
qb.multloci
would adjust whether and how many linked loci may
be considered. These apply across all chromosomes being summaried.
Returns an object of class qb.scanone
(a data frame) containing
effects selected according to type.scan
and scan
.
Brian S. Yandell, yandell@stat.wisc.edu
summary.qb.scanone
, plot.qb.scanone
data(qbExample) temp <- qb.scanone(qbExample) summary(temp) plot(temp)