st: Population averaging in panel data and applying the hurdle model in panel data

[Neil Hewitt <mcxnh4@nottingham.ac.uk>]

Dear Statalist users,

Your help on this matter would be greatly appreciated. I have searched for an answer in both the user manuals, forums, and web searches.
I am using panel data set by patient and year. I have one observation for each year my patients are on the panel but my panel dataset is unbalanced.
I have run fixed effects, random effects and population averaging.

Population averaging has got me confused, I understand pooling, but not quite how it is applied in panel.

When I have population averaged I have done so at the level of the patient. With the patient having only one observation per year is that not just the same as random effects? I have used pa, so I can achieve robust standard errors. I would prefer to use random effects over fixed effects because a number of my variables have no within group variation and are therefore lost in fixed effects. However I also suspect I have omitted variables and the hausman test comes out in favour of fixed effects. Population averaging with robust standard errors was therefore seen as a form of compromise. 

Also I have a count panel data model with an excess of zeros. My population are patients with a chronic condition and outcome a related serious hospital admission. My zeros therefore should be true zeros and I have good reason to believe that once the individual has had an event they are different to those who have not at a greater of subsequent event. Therefore the hurdle model is the most appropriate. The problem arises in applying the hurdle to the panel. I have created a binary model which predicts whether or not the patient has an event that I am happy with. The problem is with specifying the hurdle. I have said that once the patient has had an event and cleared the hurdle they remain in the count model for the remainder of their time in the panel, so do not enter the binary count model. My hurdle count model is therefore not truncated so includes zeros for years post the event if no further event took place. Does this seem a reasonable way to apply the hurdle model? Fr!
 om the applications I have seen the individual would go back into the binary component after clearing the hurdle, but this certainly does not seem sensible for my data.

Many thanks for your help,

Neil


This message and any attachment are intended solely for the addressee and may contain confidential information. If you have received this message in error, please send it back to me, and immediately delete it.   Please do not use, copy or disclose the information contained in this message or in any attachment.  Any views or opinions expressed by the author of this email do not necessarily reflect the views of the University of Nottingham.

This message has been checked for viruses but the contents of an attachment
may still contain software viruses which could damage your computer system, you are advised to perform your own checks. Email communications with the University of Nottingham may be monitored as permitted by UK legislation.





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