From brian@stat.cmu.edu  Fri Mar  5 05:59:06 1999
Received: from fervor.stat.cmu.edu (FERVOR.STAT.CMU.EDU [128.2.241.217]) by temper.stat.cmu.edu (8.6.10/8.6.6) with ESMTP id FAA03314; Fri, 5 Mar 1999 05:59:04 -0500
Received: (from brian@localhost) by fervor.stat.cmu.edu (8.6.10/8.6.6) id FAA28215; Fri, 5 Mar 1999 05:59:02 -0500
Date: Fri, 5 Mar 1999 05:59:02 -0500
From: Brian Junker <brian@stat.cmu.edu>
Message-Id: <199903051059.FAA28215@fervor.stat.cmu.edu>
To: jkau@andrew.cmu.edu
Subject: Re:  experiment
Cc: brian@stat.cmu.edu
Status: R

Good question.

The only feature that distinguishes an experiment from an observational
study is that in an experiment the researchers are responsible for who
gets what treatment.  There doesn't have to be a control group, nor
randomization, nor blinding, etc.  However without these other features
it's often difficult to learn very much from an experiment.

-BJ

>Date: Fri, 05 Mar 1999 00:50:49 -0500
>From: "John Kowalski" <jkau@andrew.cmu.edu>
>To: bj20@andrew.cmu.edu
>Subject: experiment
>
>Professor Junker,
>    What are the fewest qualities an experiment can have to be considered an
>experiment?  Just that a treatment was given to subjects?  Does there have
>to be a control group?  I used to think the former, but now I'm not certain.
>Thanks!
>-John K,