AggregateBayes {GeneSelector} | R Documentation |
The aggregated rank results from a posterior characteristic
(argument posteriorfun
below). The discrete
prior is symmetrically centered around the rank obtained
from the original dataset. The Likelihood is based on
a normal distribution with variance sigma
(s. below).
AggregateBayes(RR, S, tau, sigma = c("MAD", "sd"), posteriorfun = c("mode", "mean", "median", "quantile"), q = NULL)
RR |
An object of class RepeatRanking . |
S |
Either an object of class StabilityLm or StabilityOverlap . |
tau |
The prior variance. Controls the confidence
in the rank obtained from the original dataset. Should not be too large (<=1) in order to save computing time. |
sigma |
How the standard deviation for the Likelihood is to be estimated
from the data (=ranks from perturbed datasets). "MAD"
is a (weighted) MAD, "sd" a (weighted) standard deviation. |
posteriorfun |
Which statistic should be applied to the posterior
distribution as a summary. If "quantile" is
chosen, then it should be specified via
the argument q . |
q |
The posterior quantile used as summary statistic. Only used if posteriorfun is "quantile" |
The prior has support only in the range [r0-2*tau;r0+2*tau]
,
where r0
is the prior mode (rank from the original dataset).
The weights for the estimation of sigma
decrease linearly with
decreasing similarity of perturbed dataset and original dataset
as measured by Stability Measures (object S
).
An object of class AggregatedRanking.
Martin Slawski martin.slawski@campus.lmu.de
Anne-Laure Boulesteix http://www.slcmsr.net/boulesteix
GetRepeatRanking, GetStabilityLm, GetStabilityOverlap, AggregateSimple
## Load toy gene expression data data(toydata) ### class labels yy <- toydata[1,] ### gene expression xx <- toydata[-1,] ### run RankingTstat ordT <- RankingTstat(xx, yy, type="unpaired") ### Generate Leave-one-out Foldmatrix loo <- GenerateFoldMatrix(xx, yy, k=1) ### Get all rankings loor_ordT <- GetRepeatRanking(ordT, loo) ### compute stability measure stab_overlap <- GetStabilityOverlap(loor_ordT, decay="linear") ### aggregate rankings agg_ordT <- AggregateBayes(loor_ordT, stab_overlap, tau=1)