Date: Wed, 2 Jan 2008 20:15:30 -0500
Reply-To: Peter Flom <email@example.com>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Peter Flom <peterflomconsulting@MINDSPRING.COM>
Subject: Re: A conceptual question about regression
Content-Type: text/plain; charset=UTF-8
>Thanks for replying. My area is social science. X is the Dow Jones
>daily index; Y and Z (and some other variables like them) are certain
>characteristics in TV news broadcast later in the day. I am interested
>in the correlations between the Dow and those characteristics. As some
>suggested, is it more appropriate to run a bunch of separate
>correlations than to run one multiple regression?
Ahhhh context! There's nothing like it!
First, this is clearly a case where 'reverse' regression could work, as the news is reporting on things that happened while the Dow was trading. So, it's not really reverse in time
Second, correlations or regression? Well, what question are you trying to answer? Is it a bivariate or multivariate question?
I would tend to think bivariate regressions and then maybe a multiple regression....perhaps using GLMSELECT....
How are you classifying the 'events'?