Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

re: st: rate with count data with maximum


From   "Airey, David C" <david.airey@vanderbilt.edu>
To   "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu>
Subject   re: st: rate with count data with maximum
Date   Tue, 26 Oct 2010 17:14:16 -0500

.

After reading some more, I found that when an upper bound is present, some use a binomial regression model rather than poisson (observed/total). One example was the number of fish eggs hatched from batches of known total size. Other sources on poisson models made me think there is some confusion about variables that could be exposure/offset variables when they are not clearly time. Anyway, count models are very interesting!

> I'm using -nbreg- for count data in which I'm interested in rate, say, events per 60 minutes.
> 
> The count is made at 60 minutes for all observations, so I don't have an exposure variable.
> 
> What's different is that all observations have a maximum count that differs for each observation, that can be experimentally determined only after the observation is made, but not before. Because of the chosen time of observation, the maximum should occur after 60 minutes for most of the observations, but for a subset it may occur before. It is only known that it did occur, based on determining the maximum after the observation was made. If it occurs at 30 minutes or 59 minutes, it cannot be determined.
> 
> Typical data at 60 minutes might be:
> 
> 0/15 possible events in 60 minutes
> 1/15 possible events
> 4/15 possible events
> 12/13 possible events
> 
> etc., but also
> 
> 14/14 possible events measured at 60 minutes but possibly occurred at 30 min
> 14/14 possible events measured at 60 minutes but possibly occurred at 59 min
> 
> 10/10 possible events
> 
> I was planning on throwing out data for which the maximum had been reached within the 60 minutes, because either they started with fewer events possible (10/10 vs 14/14) or I could not determine the rate (14/14 unobserved at 30 minutes or 59 minutes).
> 
> Has anyone had count data like this?
> 
> PS...it is not known if the maximum correlates with rate.
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index