The Census Bureau guidance suggests that when you come across this problem in calculating proportions, you should use an alternative calculation for ratios. See page 19 of the attached "Accuracy" document:
Regards
From: Krishnan Viswanathan <krisviswanathan@gmail.com>
To: ctpp-news@chrispy.net
Sent: Tue, August 3, 2010 4:57:31 PM
Subject: [CTPP] Standard Error Issues
Apologies for using this forum but I figured this might be appropriate. I am trying to get the upper and lower bound at the 90% CI around shares of people who live and work in Palm Beach County and who live in PB County but work outside PB county. I am using 1 year ACS data from 2008 and am using table B08007. I am getting a negative number under the square root term and obviously it cannot calculate SE. Anyone has pointers for a workaround or what you did when faced with something like this? I doubt if it can deal with imaginary numbers and i suppose you can make it zero but if anybody can elaborate on this it that would be helpful. I am attaching the actual XLS file and you can see a screenshot of the file here: http://picasaweb.google.com/krisviswanathan/ACS_STANDARD_ERROR
TIA,
Krishnan