time series of count data...
- To: mathgroup at smc.vnet.net
- Subject: [mg83664] time series of count data...
- From: "Richard Palmer" <rhpalmer at gmail.com>
- Date: Wed, 28 Nov 2007 05:22:34 -0500 (EST)
... I have a time series of count data.. The usual tests suggest the underlying distribution is negative binomial (possibly zero inflated version). There are several potential drivers for the series (including ARMA terms of the series itself). Does anybody know a technical approach for estimating the series that deals with the count nature of the data and the autocorrelation? Does anybody have code for doing the estimation? Thanks in advance! Richard