Date:         Thu, 29 Jan 2009 17:21:22 -0500
Reply-To:     Kevin Viel <citam.sasl@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Kevin Viel <citam.sasl@GMAIL.COM>
Subject:      Re: RR estimation in case control study??

On Thu, 29 Jan 2009 12:37:28 -0500, K. C. <heeha@CANADA.COM> wrote:

>Data are collected only at the time of enrollment (sample for lab test,
>demographics and risk exposure). It's like cross-sectional.  No follow up
>is done thereafter.    Therefore, the event count per person is either 0
>(negative lab test) or 1 (positive lab test).   Enrollment is completed in
>9 months.

What might be the start time for an observational period?  One thing that
is frequently missed, in my opinion, and that Pearce emphasizes, is that
the period should always be stated.  Risk means nothing without specifying
time.  So even if you cannot construct a rate model, you may still need to
state the period.

>I have not considered proportional hazards models because I am not
>familiar with.  Would you suggest other rate models for this type of data.

The proprotional hazard model is one of the most frequently used.
Depending on whether the enrollment occurs on a rolling basis or at
intervals (the team assembles and enrolls patients), the PH model may not
be good.  At any rate you need to assemble person-time:

----x
----------o
---x
--------------x
---o

Rates are ratios of the number of events to person-time, but it is a widely
misued term (exchange rate or, worse, prevalence rate).

HTH,

Kevin