gee and bootstrapping

[Paul Mcgeoghan]

Hi,

I have a few questions:
1) I know there is syntax available for bootstrapping in regression models from Raynald's site. Can this be used for GEE models also or is there syntax for this available?
2) A project I am involved in consists of looking at the total number of financial items (out of 75 items) reported by companies on their company websites and on the central government website over a 2 year period. 3 out of 7 predictor variables vary over the 2 years, the others are constant. All predictor variables are constant over the 2 sites. As I have data for the dependent variable for 2 sites and 2 years on 74 companies, I have used a GEE model with SITE and YEAR as within subject factors and Company is the Subject variable. In YEAR 2 and on the government website SITE 2, there is very little variability as all scores are either 74 or 75 items out of 75) as it was mandatory to disclose the items on that site in the 2nd year.
Is GEE the best approach for this in SPSS 15?
How does one choose the working correlation matrix (by examining the correlation between the dependent variable scores over the 2 years and 2 sites?)
3) Before the GEE approach, I have also carried out four individual regressions for each combination of year and site individually.
As well as the total number of financial items (out of 75 items) disclosed, I also look at the total number of items for subsections (out of 15 items) for example. However, out of the 15 items, respondents give a total score of 0, 1 or 2. The distribution assumed is Normal in the GEE models, should I be using Poisson possibly when they report very few items out of the 15?
Also, in the individual regressions, is linear regression appropriate when the dependent variable scores are 0, 1, 2 out of 15?

Thanks for any advice,
Paul

==================
Paul McGeoghan,
Application support specialist (Statistics and Databases),
University Infrastructure Group (UIG),
Information Services,
Cardiff University.
Tel. 02920 (875035).

[Peck, Jon]

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Paul Mcgeoghan
Sent: Tuesday, May 29, 2007 5:37 AM
To: [hidden email]
Subject: [SPSSX-L] gee and bootstrapping

Hi,

I have a few questions:
1) I know there is syntax available for bootstrapping in regression models from Raynald's site. Can this be used for GEE models also or is there syntax for this available?
[>>>Peck, Jon]
There is a general bootstrap module available on SPSS Developer Central (www.spss.com/devcentral) that can bootstrap pretty much any SPSS output.  It requires SPSS 14 or later and the Python programmability plugin.

There is an article that may be pertinent at
http://www.ingentaconnect.com/content/asa/jabes/2001/00000006/00000001/art00006
although only the abstract is free.

HTH,
Jon Peck

Thanks for any advice,
Paul

==================
Paul McGeoghan,
Application support specialist (Statistics and Databases),
University Infrastructure Group (UIG),
Information Services,
Cardiff University.
Tel. 02920 (875035).