llcq: Log-posterior for Cq-dPCR data

In edpcr: Performs Extended Digital PCR Analysis

Description

Calculates the log-posterior or log-likelihood for the data and the parameters.

Usage

llcq(mpars, data = fetch(2, 1), maxn0 = 7, c0 = 6, 
  extra = c("trendx", "trendy", "disp", "E1"), 
  h = 0.1, n0 = 300, n1 = 465,
  nt0 = 0, nt1 = 0, n.lo = 0, n.hi = 0, 
  nx = 0, ny = 0, 
  full = TRUE, delta = 1e-05, sum. = TRUE, 
  report = 0, probreport = 0, prior = FALSE, 
  mu.fun = function(mu) 1, A.fun = function(A) 1, 
  E.fun = function(E) 1, E1.fun = function(E1) 1, 
  trendx.fun = function(x) 1, trendy.fun = function(y) 1, 
  disp.fun = function(d) 1, prior.funs = NULL, 
  invisible = FALSE)

Arguments

mpars	parameter vector in "m" parameterisation. Use conv if necessary to reparameterise.
data	cqdat object or data frame.
maxn0	the maximum number of initial molecules used in computation.
c0	number of cycles for which exact probabilities are caclulated. This can have a significant impact on speed of computation.
extra	vector of names of parameters to include in model in additon to "mu", "E", and "A". Defaults to all possibitilies: c("E1","trendx","trendy","disp").
h	threshold value Not needed if data is a cqdat object.
n0, n1, nt0, nt1, n.lo, n.hi	counts of the number of negative partitions (n0), positive partitions (n1), trimmed negative partitions (nt0), trimmed positive partitions (nt1), low outliers (n.lo) and high outliers (n.hi). Not needed if data is cqdat object.
nx, ny	numbers of rows, columns respectively.
full	logical, if TRUE then more robust method used for small log-likelihood contributions.
delta	positive number used in finite difference method.
sum.	logical, if TRUE calculate log-likelihood for all data, otherwise calculate vector of log-densities for each Cq value.
report	if postive then tracing information is produced. High values may produce more information.
probreport	if report > 0, the probability that "'" is printed when a valid log-likelihood is calculated.
prior	logical, if FALSE use constant priors, which is equivalent to returning the log-likelihood. Otherwise use priors specified by mu.fun, A.fun, E.fun, E1.fun, trendx.fun, trendy.fun and disp.fun.
mu.fun, A.fun, E.fun, E1.fun, trendx.fun, trendy.fun, disp.fun	priors as single parameter functions.
prior.funs	list of named prior functions.
invisible	logical. If TRUE then the result is printed.

Details

Due to computation limits very small log-posteriors cannot be calculated and -10^10 is returned instead. If report>0 "." will also be printed. For other log-posteriors if report>0 then "'" is printed with probability probreport.

Value

If sum. is TRUE the log-posterior, otherwise a vector of log-densities for the Cq values.

Author(s)

Philip Wilson

See Also

cqmc,plotfun

Examples

dat<-fetch(Exp37a,1)
pars0<-get.pars0(dat)
llcq(conv(pars0,"p","m"),dat)

edpcr documentation built on May 2, 2019, 5:22 p.m.