Date:         Wed, 25 Aug 1999 16:34:33 +0100
Reply-To:     John Whittington <medisci@POWERNET.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         John Whittington <medisci@POWERNET.COM>
Subject:      Re: Missing Values (... not so fast!)
Comments: To: HERMANS1 <HERMANS1@WESTAT.COM>
In-Reply-To:  <199908241749.SAA06658@mail-relay.power.net.uk>
Content-Type: text/plain; charset="us-ascii"

At 13:43 24/08/99 -0400, HERMANS1 wrote (in part):

>So, in this broader scheme of things, how would statisticians suggest that we
>represent structural zeros in frequency tables?   A missing value would,
in my
>opinion, make the problem worse.

Sig, I'm not totally certain what you mean by 'structural zeros'.  If you
are talking about the entry which should appear in a frequency table or
'crosstabulation' for a cell which is empty, then zero is surely going to
be the correct entry.  It's actually difficult to conceptualise what anyone
might mean by a 'missing' frequency!

It goes without saying (I hope!) that if one wants optimal information
presentation, any tabulation should distinguish between any
interpretationally important different reasons for absent data -
'declined', 'not applicable', 'truely random absent/loss of data' etc. -
but, again, any such 'rows/columns' in a tabulation demand numbers in them
(zero, if necessary) - 'missing' would be meaningless.

... that's how I see it, anyway!

Regards,


John

----------------------------------------------------------------
Dr John Whittington,        Voice:    +44 (0) 1296 730225
Mediscience Services        Fax:      +44 (0) 1296 738893
Twyford Manor, Twyford,     E-mail:   medisci@powernet.com
Buckingham  MK18 4EL, UK              mediscience@compuserve.com
----------------------------------------------------------------