emfit {EBarrays} | R Documentation |
Implements the EM algorithm for gene expression mixture model
emfit(data, family, hypotheses, theta.init = NULL, p.init = NULL, num.iter = 20, verbose = getOption("verbose"), trace = TRUE, optim.control = list())
data |
a matrix or an an object of class "exprSet" or "ExpressionSet" |
family |
an object of class ``ebarraysFamily'' or a characterstring which can be coerced to one. Currently, only the characters "GG" and "LNN" are valid. Other families can be supplied by constructing them explicitly. |
hypotheses |
an object of class ``ebarraysPatterns'' representing the hypotheses
of interest. Such patterns can be generated by the function
ebPatterns
|
theta.init |
initial values of theta. Should be appropriate for the corresponding
family . If omitted, defaults are calculated based on the
family .
|
p.init |
initial values of mixture probabilities. If omitted, all hypotheses are initially taken as equally likely. |
num.iter |
number of EM iterations |
verbose |
logical or numeric (0,1,2) indicating desired level of information printed for the user |
trace |
whether the result should have slots thetaTrace and
probTrace , matrices giving values for all iterations
|
optim.control |
list passed unchanged to optim for finer control
|
an object of formal class ``ebarraysEmFit'' with the following slots
|
the family passed to emfit
|
|
the hypotheses passed to emfit
|
|
numeric vector holding the parameter estimates |
|
numeric vector holding estimates of the mixture probabilities |
|
optional matrix holding intermediate values of thetaEst
|
|
optional matrix holding intermediate values of probEst
|
The show()
method can be used to summarize the object. It can
also be used to generate posterior probabilities using
postprob
emfit
is a generic function dispatched on the arguments
data
, family
and hypotheses
.
Christina Kendziorski, Michael Newton and Deepayan Sarkar
UW Madison Biostatistics and Medical Informatics Technical Report 166, ``On parametric empirical Bayes methods for comparing multiple groups using replicated gene expression profiles,'' by C.M. Kendziorski, M.A. Newton, H. Lan, and M.N. Gould. See http://www.biostat.wisc.edu/~kendzior/
ebPatterns
, ebarraysFamily-class
data(sample.exprSet.1) ## from Biobase patterns <- ebPatterns(c("1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1", "1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2")) gg.fit <- emfit(data = sample.exprSet.1, family = "GG", hypotheses = patterns, verbose = TRUE) ## theta.init = c(10, 1, 1), p.init = c(.99, .01)) show(gg.fit)