S+ 4.5 R.2, W2K,
In a bootstrap application a bootstrap sample consists of a number of
(sparse) datasets. These are pooled in order to circumvent to the low number
of observations. Within the bootstrap function I us nlmin (or nlminb) to fit
a 5 parametric curve. When the nlmin fails to converge I start a new
iteration. What puzzles me is that failure to converge comes in clusters and
in particular that the bootstrapping can enter a VERY long sequence of
unsuccessful iterations, even though the bootstrapping have started nicely.
Hope someone can explain this odd behavoiur to me and give advice as how to
fix it.
Thanks,
Rene
*****************************************
René Holst
ConStat
Groenspættevej 10
DK-9800 Hjoerring
Denmark
Tel: +45 98 92 19 79
Mob: +45 24 40 79 41
Fax: +45 98 92 19 56
Web: http://www.constat.dk
SMS: 24407941@sms.tdm.dk (max 160 characters)
*****************************************
|