adapGridProf: Profile log-likelihood of Dirichlet-multinomial model
Description
Computes the profile log-likelihood of $\ell(\pi,\theta;x)$
for an interval determined by a given difference in log-likelihood
value from the maximum log-likelihood value.
Usage
adapGridProf(data, delta, stepsize=50)
Arguments
data
A matrix or table with counts. Rows represent subpopulations
and columns the different categories of the data. Zero rows or columns
are automaticly removed.
delta
The difference between max of log-likelihood and the
profile log-likelihood. May be used to construct approximate
confidence intervals, e.g. with delta = qchisq(0.95,df=1)*2.
stepsize
The stepsize used when stepping left/right of the
MLE. The stepsize used by the algorithm is given by the MLE of theta
divided by stepsize. Default value is 50.
Value
Gives a data frame with theta values and associated profile
log-likelihood values.
data(us)
fit <- dirmult(us[[1]],epsilon=10^(-12),trace=FALSE)
adapGridProf(us[[1]],delta=0.5)
## Not run: ### adapGridProf(us[[1]],delta=qchisq(0.95,df=1)*2)