Re: st: two-step hierarchical model

[Maarten Buis <maartenlbuis@gmail.com>]

On Sun, Nov 3, 2013 at 10:15 PM, Amilcar Manuel Reis Moreira wrote:
> I am using a two-step hierarchical model to look at the factors that influence the likelihood of retired persons supporting cuts to spending on unemployment benefits. I have a database that includes data for 20 nations, in two different points in time: 1996 and 2006.
>
> In the first step, I use a multinomial logistic model to estimate the predicted probability  that retired persons support cutting expenditure on unemployment benefits, in each country/year.
>
> My aim is , in the second step, to estimate the impact of the state of the economy (measured by the unemployment rate) and the design of unemployment benefits system (assistance vs. insurance based and generosity on the predicted probability  that retired persons support cutting expenditure on unemployment benefits, in each country/year. Given the limitations of my database, what would be the best method to estimate this?

Such advise depends on detailed knowledge of the _exact_ limitations
of the database and _exactly_ what you want to do. So to give a good
answer we need much much much more information. However, internet
forums are not a good way to communicate such detailed descriptions.
This is much better done through face-to-face communication. Forums
like Statalist work best for very narrow specific problems. So the
best advise we can give you is to find someone local with which you
can meet and discuss your research design.

Some small hints can be given. Since you want to do a multinomial
logit, I suspect you have a dependent variable in the form a a Likert
like item: the respondent is given a statement like "I support cuts in
umemployment benefit" and can answer whether (s)he agrees a lot, a
bit, not so much, or not at all with that statement. In that case your
dependent variable is ordinal and Stata does have methods for dealing
with multilevel ordinal data, see: -help meologit-. However, there is
some evidence that you need to be careful with few higher level units
(in your case countries). See for example this talk by Stephen Jenkins
at the last UK Stata Users' meeting:
http://www.stata.com/meeting/uk13/abstracts/materials/uk13_jenkins.pdf

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------

*
*   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/