Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: clogit and autocorrelation


From   "Nick Cox" <n.j.cox@durham.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: clogit and autocorrelation
Date   Fri, 2 May 2008 13:21:09 +0100

It sounds as if you need -xtclogit-, but I don't think it exists. 

I guess your problem is thus that the kind of response variable you have leads to a model that doesn't have support for panel data. 

Durbin-Watson tests won't be available after -clogit-. Rather than that, two not very confident suggestions are

calculating residuals and looking at their autocorrelation functions and e.g. a plot of residual vs previous residual 

considering whether lagged predictors would be useful extra predictors within -clogit-. 

That is, you can -tsset- your data and use operators like L. to do manipulations. But that is outside your logit modelling. -clogit- will do nothing special with panel data. 

There is probably better advice out there. 

C.Chetkiewicz

I have a matched 1:20 dataset of 4 hour telemetry steps that I am analyzing in Stata using clogit.  I am clustering on individual animal to produce robust SEs.  However, I want to look at the autocorrelation structure of the data and have never done this before.  I am using only non-collinear variables in my model.
  
1. How do I explore autocorrelation and partial autocorrelation of model residuals with matched data.   
2. I assume I can use the Durbin Watson test for temporal autocorrelation but is this a straight call after clogit or do I need to reformat the dataset specifically for time series?  Right now I have a generic hour and generic date in my work. 


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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   |   What's new   |   Site index