Dear Clarify Support,
I would be very grateful if you could offer advice on the following problem.
I cannot figure out how to get clarify to return expected values for a simple
model with an interaction term.
I am modeling government responsiveness (continuous, [0,1]) as a function of a
competitive electoral environment (a dummy, 1=competitive) and literacy
(continuous, [0,1]). Each observation is a municipality. I expect both
variables to have a positive effect on responsiveness; but I also expect
literacy to have a greater effect when coupled with electoral competition, so
I interact the two variables.
The Clarify documentation explains how to hold this interaction term at its
median, while simulating other values of interest. But I need to manipulate
the interaction term itself. I'm trying to produce two variables: one is the
expected value of responsiveness when competition=0, and another when
competition=1.
For the first variable, the setx command provides an easy solution:
local a=1
while `a'<=99 {
setx competition 0 literacy p`a' interaction 0
simqi ....
genev ....
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
Dear Clarify Support,
I would be very grateful if you could offer advice on the following problem.
I cannot figure out how to get clarify to return expected values for a simple
model with an interaction term.
I am modeling government responsiveness (continuous, [0,1]) as a function of a
competitive electoral environment (a dummy, 1=competitive) and literacy
(continuous, [0,1]). Each observation is a municipality. I expect both
variables to have a positive effect on responsiveness; but I also expect
literacy to have a greater effect when coupled with electoral competition, so
I interact the two variables.
The Clarify documentation explains how to hold this interaction term at its
median, while simulating other values of interest. But I need to manipulate
the interaction term itself. I'm trying to produce two variables: one is the
expected value of responsiveness when competition=0, and another when
competition=1.
For the first variable, the setx command provides an easy solution:
local a=1
while `a'<=99 {
setx competition 0 literacy p`a' interaction 0
simqi ....
genev ....
.
.
local `a'=a+1
}
This returns the expected value of the d.v. for each percentile of literacy,
when competition=0.
However, for the second part of the problem, when competition=1, I can think
of no way to use setx appropriately. I can't set the interaction term at its
own percentiles, since the value is 0 for many cases (having been interacted
with a 0 from the competition variable), while the whole point of the second
simulation is to set competition at 1. What I would like to do (I think) is
to set the interaction variable according to the percentiles of literacy; but
there is no way in the program to setx the values of one variable based on the
percentiles of another.
If you have any suggestions as to how I can resolve this problem, I would
appreciate it. I've been using Clarify for my dissertation research, and it
has really helped with the interpretation and presentation of my results.
Best,
Matt Cleary
University of Chicago
--
Clarify mailing list served by Harvard-MIT Data Center
[Un]Subscribe/View Archive: http://lists.hmdc.harvard.edu/?info=clarify
