Statistical testing tool question

In the Census Bureau’s Excel Statistical Testing Tool, the sample margins are preceded by “+/-.” Does the tool still calculate statistical significance correctly without the “+/-“ signs? I have to manually enter many of these margins (rather than simply copying from data downloads), and Excel will simply not allow me to enter those characters before numbers. It keeps forcing me to correct them.

www.census.gov/.../statistical-testing-tool.html

I don’t see documentation from the Census Bureau instructions that specifically address this. The “worked example” spreadsheet in the Testing Tool file includes an example with parentheses around the margins, which Excel will allow, but I’d prefer to keep the editing to a minimum – i.e., it would be easier if the +/- is not really necessary to calculate significance.

I’ve also tried copying and removing the +/- symbols and the statistical significance results appear to come out the same for a few examples, but I’d prefer to have some authoritative confirmation of this. Thanks!

  • Not an authoritative answer, but I've used the tool without +/- and have not had problems with it.
  • I think you leave out the +/- sign. As far as I know, the math equation the tool uses is:
    absolute value of estimate1 - estimate 2 / square root of MOE1^2 + MOE2^2
    Since the margins of error are squared in the equation, whether or not there's a plus or minus sign in front of them is moot!
  • I talked to my colleague who developed the tool. He said:

    Yes, the tool should work properly. The data users are free to unprotect the sheet and examine the equations. They follow a number of nested if statements, and I'm not sure it is the most easily to follow. However, it is designed to look for special cases that you might see if you are copying MOEs directly from AFF (when the tool was developed). However, if the MOE is just a number, it shouldn't be a problem.

    In addition, there are some parameters at the very bottom of the 2 stat. testing tabs that allow data users to change to using standard errors instead of MOEs (or change to using a 95% confidence level MOE, for example).
  • In reply to Gretchen Gooding:

    Beth, Bill, and Gretchen,
    Thanks to all of you for the clarification! This is very helpful.
    Mike