The correction factor \(\log(q/r)/r\) for the likelihood root is unbounded in the vincinity of the maximum likelihood estimator. The thesis of Rongcai Li (University of Toronto, 2001) explores different ways of bridging this singularity, notably using asymptotic expansions.
tem.corr(fr, print.warning = FALSE)
an object of class fr
, containing as additional arguments spline
and a modified rstar
argument.
an object of class fr
logical; should warning message be printed? Default to FALSE
The poor man's method used here consists in fitting a robust regression to \(1/q-1/r\) as a function of \(r\) and using predictions from the model to solve for \(q\). This approach is seemingly superior to that previously used in spline.corr.