Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down at the end of May, and its replacement, statalist.org is already up and running.


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

Re: st: Question: Ordered logit- suspicious odds ratio


From   Richard Williams <richardwilliams.ndu@gmail.com>
To   statalist@hsphsun2.harvard.edu, statalist@hsphsun2.harvard.edu
Subject   Re: st: Question: Ordered logit- suspicious odds ratio
Date   Mon, 16 Apr 2012 10:27:48 -0500

At 06:12 AM 4/16/2012, Stefanie Kneer wrote:
Dear Statalist

I am encountering running an ordered logit regression with the
following panel data ranging from 1992-2010.

describe  p11101 regionunempl unempl_empl interaction



storage  display  value

variable name   type   format label variable label



p11101          byte   %8.0g        life satisfaction -discrete from 0-10

regionunempl    float  %9.0g        Region unempl. %

unempl_empl float %9.0g dummy for being 1=unemployed and 0=employed

interaction float %9.0g regionalunemployment rate*unemployment dummy

I am trying to identify the link between unemployment and happiness
and in particular whether there exists a social norm effect. P11101 is
thus my dependent variable and as it is discrete I was planning to run
an ordered logit but I am encountering difficulties when trying to
calculate the odds ratio. As can be seen the social norm effect, and
thus whether you feel more comfortable when other people around you
are not working, too is massively big (27). That looks kind of
suspicious to me.  Could you give me an advice on why this is the case
and what I should do about it?

I think it is hard to say that an effect is "massively big" without knowing something about the scaling. If you rescale the variables, e.g. measure income in dollars rather than 1000s of dollars, the coefficients will get rescaled too. If regionunempl is coded on a scale that ranges from 0 to 1, then a 1 unit increase is a HUGE increase (100 percentage points), indeed probably a much bigger increase than actually occurs in your data. If it is coded 0-100, then a 1 unit increase is just a 1 percentage point increase, which is much smaller. So, consider rescaling your variables so a 1 pt increase is a meaningful number. Or, compute the predicted values, or use -adjust- or -margins- to compute predicted values for various combinations of values, and see if the results seem plausible to you.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index