Re: st: group difference

[Maarten Buis <maartenlbuis@gmail.com>]

On Sun, May 26, 2013 at 10:05 PM, David Hoaglin wrote:
> Education level and gender composition seem very likely to be
> categorical, and income may be categorical.  A t-test would usually
> not be appropriate.
>
> To make matters more interesting, the categories of education level
> (and income, if categorical) are ordered.  The -ologit- command with
> race/ethnicity as the predictor would be a good place to start.

To add to the complexity: education could be measured as a (pseudo)
continuous variable: years in school or "typical" number of years
needed to achieve the respondent's achieved diploma. On the other
hand, in tracked educational systems the ordering of different
outcomes can be ambiguous. You could write a complete PhD thesis on
the subject of measuring education and in fact somebody has already
done that: (Schneider 2009).

This is just to reinforce David's comment that you need to look at how
exactly these variables were measured and what they measure exactly:
What was the exact question (or questions) on which this variable is
based? What possible answers could the respondent give? If it was
categorical: which categories? Imagine you were asking this question
to your parrents/your neighbour/..., would they understand this
question and the answer categories? For example, the answer categories
for an education question typically correspont with the current
educational system, which, due to educational reforms, can be very
different from the educational system in which respondent attained his
or her education. How were these answers transformed to the values in
your variable?

Hope this helps,
Maarten

Schneider, Silke L. (2009): Confusing credentials: the
cross-nationally comparable measurement of educational attainment,
DPhil thesis. Oxford: University of Oxford.

---------------------------------
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/