Date: Mon, 6 Aug 2007 16:11:20 -0700
Reply-To: Richard <richard.hockey@GMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Richard <richard.hockey@GMAIL.COM>
Subject: Re: PHREG and big datasets
Content-Type: text/plain; charset="us-ascii"
On Aug 6, 11:15 pm, plessthanpointohf...@gmail.com wrote:
> Hi Ben,
> This is stored medical data from the past 15 years from a source that
> shall remain nameless at this time. We have permission to access it
> but we don't own it, or the servers it's on, etc. The computers we
> use should be updated soon and I am hoping that will alleviate much of
> the problem. However, I will never be able to bring this over to
> unix. PC will be my only available processing tool...
> I got a suggestion from Melvin at SAS-L to try running samples of
> different sizes and see what happens to processing time. So I am
> going to start there.
Looking at your sample data I would say that you first need to flatten
your data. It seems to have multiple records for each person but you
are treating these as independent observations. You also have time
dependent covariates which need to be explicitly defined in the PHREG
program lines. How many individuals are in your dataset? You may find
after flattening that the number is quite manageable.