The interaction is evident in the greater rightward shift of the old Catholics
from the young Catholics than of the old non-Catholics from the young non-
Catholics.
Don't expect these lines to cross as evidence of an interaction. A difference
in the slopes of the lines of indicates heteroscedasticity and not necessarily
interaction.
These plots are analogous to dose-response function plots in pharmacology: to
interpret them, draw a horizontal line (parallel to the x axis) through all of
the plotted lines. Choose a value that allows the horizontal line to intersect
all of the plotted lines; in this example -yline(0.5)- works. The distance
along the horizontal line between points of intersection is a measure of
proclivity to attend. Difference in the lengths along this line, i.e.,
difference in this proclivity as a function of age, is greater among Catholics--
that's the interaction.
Joseph Coveney
--------------------------------------------------------------------------------
clear
set more off
input str18 group str6 attendance int freq
"young non-Catholic" "low" 322
"young non-Catholic" "medium" 122
"young non-Catholic" "high" 141
"old non-Catholic" "low" 250
"old non-Catholic" "medium" 152
"old non-Catholic" "high" 194
"young Catholic" "low" 88
"young Catholic" "medium" 45
"young Catholic" "high" 106
"old Catholic" "low" 28
"old Catholic" "medium" 24
"old Catholic" "high" 119
end
label define Attend 1 low 2 medium 3 high
encode attendance, generate(attend) label(Attend)
compress
version 7: ordplot attend [w=freq], by(group) ///
sc(logit) reverse pla(0.1(0.1)0.9) xla(1/3) yline(0.5)
exit
--------------------------------------------------------------------------------
Janet,
I think you are correct. The "indication of an interaction" is quite weak
if it exists at all. I suspect an error might have been made originally
when the data was set up...I used -encode- on the "attend" variable, and
it coded differently than the low/medium/high = 1/2/3 suggested by the
help text, and as a result I did observe an interaction using the
incorrectly coded data...perhaps; there is also weak suggestion of an
interaction if the plot is on the raw scale rather than on the logit
scale.
Anyway, maybe you are missing something obvious, but if you are, so am I
Buzz Burhans
> I have been using ordplot and thought that I understood it until I tried
> the
> Knoke and Burke example in the help file which states that the resultant
> plot shows evidence of an interaction. I have been looking at this plot
> for
> a couple of days and I cannot see the interaction. I realise that I am
> missing something blindingly obvious. Can somebody please enlighten me.
> Thanks,
> Janet
>
*
* 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/
