summary.scantwo {qtl} | R Documentation |
Summarize the interesting aspects of the results of scantwo
.
## S3 method for class 'scantwo': summary(object, thresholds, what=c("best", "full", "add", "int"), perms, alphas, lodcolumn=1, pvalues=FALSE, df=FALSE, allpairs=TRUE, ...)
object |
An object of class scantwo , the output of the
function scantwo . |
thresholds |
A vector of length 5, giving LOD thresholds for the full, conditional-interactive, interaction, additive, and conditional-additive LOD scores. See Details, below. |
what |
Indicates for which LOD score the maximum should be reported. See Details, below. |
perms |
Optional permutation results used to derive thresholds or
to calculate genome-scan-adjusted p-values. This must be consistent
with the object input, in that it must have the same number of
LOD score columns. |
alphas |
If perms are included, these are the significance levels used
to calculate thresholds for determining which peaks to pull out. It
should be a vector of length 5, giving significance levels
for the full, conditional-interactive, interaction, additive, and
conditional-additive LOD scores. (It can also be a single number, in
which case it is assumed that the same value is used for all five LOD
scores.) If thresholds is specified, alphas should not
be. |
lodcolumn |
If the scantwo results contain LOD scores for multiple phenotypes, this argument indicates which to use in the summary. Only one LOD score column may be considered at a time. |
pvalues |
If TRUE, include columns with genome-scan-adjusted
p-values in the results. This requires that perms be
provided. |
df |
If TRUE, the degrees of freedom associated with the LOD scores are shown. |
allpairs |
If TRUE, all pairs of chromosomes are considered. If FALSE, only self-self pairs are considered, so that one may more conveniently check for possible linked QTL. |
... |
Ignored at this point. |
If what="best"
, we calculate, for each pair of chromosomes, the
maximum LOD score for the full model (two QTL plus interaction) and
the maximum LOD score for the additive model. The difference between
these is a LOD score for a test for interaction. We also calculate
the difference between the maximum full LOD and the maximum single-QTL LOD
score for the two chromosomes; this is the LOD score for a test for a
second QTL, allowing for epistasis, which we call either the
conditional-interactive or "fv1" LOD score. Finally,
we calculate the difference between the maximum additive LOD score and
the maximum single-QTL LOD score for the two chromosomes; this is the
LOD score for a test for a second QTL, assuming that the two QTL act
additively, which we call either the conditional-additive or "av1" LOD
score. Note that the maximum full LOD and additive LOD are allowed to
occur in different places.
If what="full"
, we find the maximum full LOD and extract the
additive LOD at the corresponding pair of positions; we derive
the other three LOD scores for that fixed pair of positions.
If what="add"
, we find the maximum additive LOD and extract the
full LOD at the corresponding pair of positions; we derive
the other three LOD scores for that fixed pair of positions.
If what="int"
, we find the pair of positions for which the
difference between the full and additive LOD scores is largest, and
then calculate the five LOD scores at that pair of positions.
If thresholds
or alphas
is provided (and note that when
alphas
is provided, perms
must also), we extract just
those pairs of chromosomes for which either (a) the full LOD score
exceeds its thresholds and either the conditional-interactive LOD or
the interaction LOD exceed their threshold, or (b) the additive LOD
score exceeds its threshold and the conditional-additive LOD exceeds
its threshold. The thresholds or alphas must be given in the order
full, cond-int, int, add, cond-add.
Thresholds may be obtained by a permutation test with
scantwo
, but these are extremely time-consuming.
For a mouse backcross, we suggest the thresholds (6.0, 4.7, 4.4, 4.7,
2.6) for the full, conditional-interactive, interaction, additive, and
conditional-additive LOD scores, respectively.
For a mouse intercross, we suggest the thresholds (9.1, 7.1, 6.3, 6.3,
3.3) for the full, conditional-interactive, interaction, additive, and
conditional-additive LOD scores, respectively. These were obtained by
10,000 simulations of crosses with 250 individuals, markers at a 10 cM
spacing, and analysis by Haley-Knott regression.
An object of class summary.scantwo
, to be printed by
print.summary.scantwo
;
The previous version of this function is still available, though it is
now named summary.scantwo.old
.
We much prefer the revised function. However, while we are confident
that this function (and the permutations in
scantwo
) are calculating the relevant statistics,
the appropriate significance levels for these relatively complex
series of statistical tests is not yet completely clear.
Karl W Broman, kbroman@biostat.wisc.edu
scantwo
, plot.scantwo
,
max.scantwo
, condense.scantwo
data(fake.f2) fake.f2 <- calc.genoprob(fake.f2, step=5) out.2dim <- scantwo(fake.f2, method="hk") # All pairs of chromosomes summary(out.2dim) # Chromosome pairs meeting specified criteria summary(out.2dim, thresholds=c(9.1, 7.1, 6.3, 6.3, 3.3)) # Similar, but ignoring the interaction LOD score in the rule summary(out.2dim, thresholds=c(9.1, 7.1, Inf, 6.3, 3.3)) # Pairs having largest interaction LOD score, if it's > 4 summary(out.2dim, thresholds=c(0, Inf, 4, Inf, Inf), what="int") # permutation test to get thresholds; run in two batches # and then combined with c.scantwoperm ## Not run: operm.2dimA <- scantwo(fake.f2, method="hk", n.perm=500) operm.2dimB <- scantwo(fake.f2, method="hk", n.perm=500) operm.2dim <- c(operm.2dimA, operm.2dimB) ## End(Not run) # estimated LOD thresholds summary(operm.2dim) # Summary, citing significance levels and so estimating thresholds # from the permutation results summary(out.2dim, perms=operm.2dim, alpha=rep(0.05, 5)) # Similar, but ignoring the interaction LOD score in the rule summary(out.2dim, perms=operm.2dim, alpha=c(0.05, 0.05, 0, 0.05, 0.05)) # Similar, but also getting genome-scan-adjusted p-values summary(out.2dim, perms=operm.2dim, alpha=c(0.05, 0.05, 0, 0.05, 0.05), pvalues=TRUE)