-----Original Message-----
From: Paul Brandon [mailto:[EMAIL PROTECTED]]
Sent: Monday, February 26, 2001 3:50 PM
To: [EMAIL PROTECTED]
Subject: Re: Gallup/creationism
At 1:34 PM -0600 2/26/01, Jim Guinee wrote:
>> Paul Smith wrote:
>>
>> To his credit, when he announced the faith-based program office,
>> Bush explicitly said that the goal was to only promote programs that had
>> supporting empirical evidence. An interesting contradiction,
>
>Why? What about a faith-based program that has demonstrated efficacy in
>solving a particular social problem (promoting abstinence to decrease std's
>and unwanted pregnancies)?
Paul Brandon writes:
"Do you know of any? Most have highly selected admissions, as well as
excluding dropouts from the data anaysis."
Are these problems limited to faith-based programs or are they common to
many social program evaluations? The highly selected admissions may be a
problem with external validity but it doesn't mean that the program would
not be effective in other situations in which the admissions process is
equally selective. Even medical studies have certain limitations as to who
can participate.
As far as excluding dropouts from analysis, mortality is always a troubling
threat to internal validity but I am not aware of a good solution for it. If
the treatment group's mortality differs significantly from the control
group, it may call the results into question. However, how would you include
dropouts in the data analysis if your DV is more sensitive than just
success/failure? You can look at it after the fact and see if the groups
differ in mortality but there is not much to do to correct it. BTW, I am not
being rhetorical. If there is a good methodological or statistical solution
for mortality, I would like to share it with my research methods class.
Thanks,
Rick
