Speed of models, esp. pcountOpen

46 views
Skip to first unread message

Daniel Hocking

unread,
Jul 1, 2013, 9:25:11 AM7/1/13
to unma...@googlegroups.com
When I increase the range of values to be integrated over (K) in pcountOpen, the time to run the model increases exponentially. I know it's best to try the smallest value of K that makes biologically sense as a maximum and results in model convergence. I am just curious if anyone know why time increases exponentially rather than linearly? Is it just that convergence on the maximum likelihood gets less stable when including the area under the far tail end of the curve? I haven't spent much time coding my own ML functions, so I'm not really familiar with the these issues, just curious.

Thanks,
Dan

Jeffrey Royle

unread,
Jul 1, 2013, 9:29:50 AM7/1/13
to unma...@googlegroups.com
hi Dan,
 I believe (but am not certain about this) that there are double summations involved in the likelihood evaluation so the number of calculations increases at a K^2 rate or faster maybe.  I vaguely remember that from the original Dail and Madesen paper (which I don't have handy here).
regards
andy
 



--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

Reply all
Reply to author
Forward
0 new messages