Janet Thurgood wrote:
> Dear Buzz and Joseph,
> Thank you both for your assistance. I now understand the interpretation of
> the plot. I admit that I expected lines to cross to indicate an interaction.
> I will have to go back to the original references to improve my
> understanding.
Interpreting an indication of an interaction in a nonlinear system is tough.
(There's an article on the topic of interaction in selected nonlinear models in
the current issue of _The Stata Journal_.) In the example of Knoke and Burke
from the helpfile for -ordplot-, interpretation of the differences in degree of
rightward displacement (along a scale made up of an ordered categorical
variable) as interaction would have been more straightforward had the lines for
the young (or for the old) of the two denominations been superimposable.
Modeling might help sometimes. In the current example, it might be worthwhile
to construct a model by creating dummy variables for age group and
denomination, forming an interaction term, and using, for example, -ologit-
(proportional odds) or one of the alternative, user-written commands available
in Stata for ordered categorical data. A snippet of code along this line that
can be tacked onto the earlier posting's do-file is attached below, and its
results can be construed as providing supportive evidence for what's intimated
in -ordplot-. Again, the article in _The Stata Journal_ has much more to say
in this complicated area.
Joseph Coveney
generate byte young = index(group, "young") > 0
generate byte catholic = index(group, "non") == 0
xi: ologit attend i.young*i.catholic [fweight = freq]
estimates store A
xi: ologit attend i.young i.catholic [fweight = freq]
lrtest A .
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
