How to expand the change-point detection analysis to multiple time periods (i.e. days, hours)?

Wondering how you deal with the fact that the time period you found using the methodlogy might show a cosniderable different distribution parameter setting on the next day or week.

11 November 2012 6,501 0 View