Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

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

From |
Nicola Man <n.man@unsw.edu.au> |

To |
"statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |

Subject |
Re: st: question on zero inflated regression |

Date |
Mon, 21 Feb 2011 14:55:56 +1100 |

Hi Rachel, Yes, you are quite likely to be correct about variables not measured for the zero-inflated component. It was a relatively small data set that has few variables measured and I had actually been advised to take the Bayesian approach instead which I am still working on at the moment. Regards, Nicola ------------------------------ Date: Thu, 17 Feb 2011 10:07:29 +0000 From: rachel grant <rachelannegrant@gmail.com> Subject: Re: st: question on zero inflated regression Thanks Nicola and Nick for your input. I suppose another possibility in your case Nicola that there are variables generating the zeros (no parasite eggs) but you have not measured these? In my case, two of seven variables were significant predictors of zeroes (both temperature), and that makes sense because at low temperatures amphibians cannot move. I also have checked the Voung score and it indicates that zero inflated is better for my data. I have two other questions however! 1. If I change the order of the variables sometimes the p value of each variable changes although the overall LR and P for the model remains similar. Why does this happen and what can i do about it? 2. when I try to get a better fitting model by removing nonsignificant variables the log-likelihood decreases slightly. I am not sure why this happens or what to do Many thanks Rachel Grant Dept. Life Sciences Open University UK On 17 February 2011 07:26, Nicola Man <n.man@unsw.edu.au> wrote: > I have a different take on the decision to drop the zero-inflated model if there are no variables predicting excess zeros. For example, in parasite egg count data which I have been working on, it is known that generally a zero-inflated model may fit better. I might have no predictors for the excess zeros, but the zero-inflation model still works better because the fit to the neg bin distribution is better than that of the standard model. I would use the vuong test to test whether the zero-inflated model is better than the standard model for this decision on which distributional model to take. > > Regards, > Nicola > > ------------------------------ > > Date: Sun, 13 Feb 2011 13:41:13 -0800 > From: Owen Gallupe <ogallupe@gmail.com> > Subject: Re: st: question on zero inflated regression > > Hi, > > Seems to me that there are a number of ways you can go with this: > > 1) if you don't have a theoretical justification as to why there is an > inflation of zeros, put in all variables as potential causes. > 2) don't use the zero inflated model at all. If there is reason to > suspect that the zero scores are simply part of the continuum of > scores (as in, the same causal factors related to scores of 1, 2, > etc., are the same factors as scores of zero but in varying degrees), > then a standard negative binomial might be better. > > In my opinion, this is a theoretical question. > > Owen Gallupe > > > On Sun, Feb 13, 2011 at 10:58 AM, rachel grant > <rachelannegrant@gmail.com> wrote: >> Hi I am new to the list and pretty clueless. >> I am trying to use Zero inflated models, my data are counts of >> ampibians arriving at a breeding site per night. Most nights none >> arrive or one, but some nights many arrive hence the data are >> overdispersed and have excess zeroes. I am using zero inflated >> negative binomial regression. What I am confused about is I have to >> specify the inflation variable (ie the one that is generating the >> excess zeroes). I have 7 predictor variable and have no clue which is >> responsible for generating the zeroes, so what to put into the model >> as the inflation variable? Thanks for reading this >> >> -- >> regards, Rachel >> >> Rachel Grant >> Dept. Life Sciences >> Open University >> UK >> >> >> >> >> -- >> regards, Rachel >> >> * >> * 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/ >> > > * > * 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/ > > ------------------------------ > > * > * 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/ > - -- regards, Rachel * * 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/ * * 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/

- Prev by Date:
**st: calculating value weighted portfolio returns** - Next by Date:
**st: svy adjustment for rates** - Previous by thread:
**Re: st: question on zero inflated regression** - Next by thread:
**st: SIPP survey in stata - calculations** - Index(es):