> From: David N. Beede [mailto:David.Beede@MAIL.DOC.GOV] > I face the ugly possibility of having to process what I > consider to be a lot of > hard copy data (about a one-foot stack of paper). Almost all > of it is typed > (i.e., not hand-written) but in different tabular formats. > As I see it, I can > either scan in the data and write a SAS program to get the > data into shape; I > can write a program (SAS? Excel?) to create an on-screen > form to facilitate > keypunching the data; or I can contract out the keypunching.

many different tabular formats: go keypunch

> I assume others have or will face this kind of problem. My > questions are:

> -- is scanning more trouble than it is worth?

Volume is the determining factor. Just like with a press run: once you do the set-up -- the expensive part -- extra copies are just the cost of running the press. Writing a scanning program is _not_ an ad hoc process. Most scanning packages will provide both a method of designing the data collection form and validating the data. Register is key when scanning a document, just getting the four corners of the page lined up so we know where they are before we start the work: we can't have them suckers running in thru the scanner cock-eyed, y'know. ;-)

> -- does scanning require additional software other than what > comes with the scanner?

Surprise: you get get either bit-maps or text! Be prepared to spend a whole bunch of time -- remember the 80/20 rule -- reviewing the data coming off the paper.

My blue-sky approach to this has always been to suggest that folks scan twice: once at high-resolution, meaning slow reading time, high accuracy again at low-resolution, meaning fast reading time, low accuracy then use proc COMPARE to find differences. My visionary approach hasn't been implemented, because of contracting of data transfer. <sigh> well, under-appreciation happens! LOL

> -- are there any SAS-related issues in processing scanned data?

You better have your data dictionary written, and not chiseled in stone, before you begin this project.

> -- are there any published references on how to create forms > for keypunching data?

http://www.google.com scanning

Ron Fehd the macro maven CDC Atlanta GA USA RJF2@cdc.gov ---> cheerful provider of UNTESTED opinions !*! <---