Hi Melissa,
I came late to the party and reacted to the content of John's response
without having read the earlier elements of the thread. Since you
responded to my post I read the whole thread and see that you had
already unambiguously defined what you needed regarding data entry.
An aspect that seems to have been lost in the responses is that you
are vulnerable to audits for some span of time after the data entry.
To that end, I would not delete any data entered by anyone until the
times for audits have long since passed. While accuracy of the data
is the intent, being able to prove that you did it all according to
the terms of the grant are your ultimate fallback. The best way to
prove that you had more than one data entry stream is to retain all of
the data streams.
To get more specific and applicable guidance you'll have to
communicate the business rules for the whole process. If those rules
haven't been given to you then you'll have to make them up by
analyzing the requirements and designing the solutions. How will you
identify a particular user and data stream? Will each user & stream
combination eventually be locked from further editing? How long can a
user return to edit the stream? Can a supervisor edit someone else's
stream? (not a good idea. better to have the supervisor create
her/his own audit stream).
Apparently some significant amount of thought has already gone into
the process as you say that there are 20 tables. Knowing the
structure of your data might also help.
FWIW I would keep all of the data streams and the production data in
the same set of tables. There should be an attribute within the
"main" record that identifies its provenance. With proper
relationships, only that "main" record needs that attribute. There
can be a separate table that holds the values of that attribute such
as "Audit", "Entry" and "Production". Per relational rules, all
entities of the same type belong within the same table. To separate
entities into different tables based on some attribute is not a good
idea. It's tantamount to a car dealerships having separate tables in
their databases for each color of car: tblRecCars, tblBlueCars,
tblWhiteCars, etc. Color is obviously an attribute and its value
should appear in a field in a record.
It seems that you'd like is for someone to come up with a good process
for the most complex part of the whole process: Comparing the data
streams and resolving data conflicts by selection (and probably
further editing) while simultaneously referring to the input data
documents. That can be done but it requires a very good understanding
of your entire process, its data structures and workflows.
HTH
--
-Larry-
--
mlthomas007 via AccessMonster.com said:
Hi,
The requirements for dual data entry are imposed so two persons will enter
the same data to be error checked. They will use the same hard copy source
to do so. The study grant from the IRB has been doing this for years but the
issue we are running into is how to set this up best in access. Using an
access database that has already been developed without being normalized.
Thanks,
Mel
Larry said:
Mel,
I would verify exactly what the words in the requirement for "dual
data entry" mean. I suspect that the intent is that there be two
entirely separate data entry streams of each data element.
Going on the above assumption:
Neither person entering the data should know or care about anyone else
entering the same data. There should especially be no "steering" of
the data that would occur if the person entering the "later" data
stream were constantly influenced by what had been entered in the
parallel, "first", data stream. In other words, I'd assume that the
intent was that totally independent operators view and interpret the
input data and enter the data as they, independently, interpret it to
be. There should be no bar to having more than one data stream being
entered simultaneously by separate typists. The validating or
cross-checking process should not begin until all data streams of the
same data have been completed. It should be part of the checking and
resolution process to display the elements that disagree and to
resolve them by some designed process. The resolution process should
be specified as part of the over-all application.
The long and the short of it is that the business requirements must be
clearly defined before the technical solutions are designed and
implemented.
HTH
[quoted text clipped - 11 lines]