st: RE: Cross sectional analysis with ordered probit

[Eilya Torshizian <e.torshizian@auckland.ac.nz>]

Hi Andre,

Yes you can have a pooled cross section, however you cannot do fixed effects. One way is to use random effects. Another is to use Synthetic panel by creating some groups based on specific characteristics. I assume you have some weights as it is a survey. So you should use them in your estimation later on.

You probably can use 'ttest' for doing a mean test amongst different waves.

Based on categorical variables' theory, you should assume a base group in your estimation. For example, younger people who are less happy in year 1982 can be a base group to be compared with other groups. After taking this comparative approach you will not have the collinearity problem.

You can find an example of Probit and interpretations here, http://www.ats.ucla.edu/stat/stata/dae/probit.htm .

Regards,
Eilya.

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Andre, Sophie
Sent: Thursday, 8 August 2013 8:05 a.m.
To: Ben Krauss; statalist@hsphsun2.harvard.edu
Subject: st: Cross sectional analysis with ordered probit

Dear Statalisters, 

I am novice in using this software and I am a bit lost regarding the implementation of the model I run. I am doing a cross sectional analysis over 85 countries using ordered probit model. I would like to observe the interactions between micro and macro variables.

1) My first question is more econometric related. I am using World Values Survey which report data for the period 1981-2008. In this sample, some years are missing, data comes from different individuals and countries each year and the number of individuals and countries varies each year. Can I use a pooled cross section? Does it mean I only have to run one regression with years fixed effect? 

2) So as to use a pool cross sectional analysis, the temporal stability of my variables is required. It is however not possible to run a Chow test for testing the temporal stability of my variables since my model is non-linear models. Is it then ok to run it with OLS? It is indeed well-known in my subject that OLS and Ordered probit regression give the same results when regressing happiness. 

3) When running the following regression:
oprobit happiness GDPlog inc_scl GDPlogXinc_scl trust_trait age collectivism_bin sex development year_1982 year_1984 year_1989 year_1990 year_1991 year_1995 year_1996 year_1997 year_1998 year_1999 year_2000 year_2001 year_2002 year_2003 year_2004 year_2005 year_2006 year_2007 year_2008 scandinavia former_soviet_union eastern_europe latin_america asia africa the_balkans middle_east

I have some year dummies variables that are omitted as well as the interaction term GDPlogXinc_scl (log GDP x personal income scale) because of collinearity. How can I solve this problem since I need these variables to be taken into account?

4) Finally, I have big doubts regarding the interpretation of the coefficient of the dummies variables in ordered probit model. Could you please provide me an example?

I am really sorry for all these questions but the lack of experience with Stata is a problem when dealing with ordered probit models but I really would like to do things correctly this model. 

Thanks very much in advance, 
Sophie

*
*   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/

*
*   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/