# Statistics: Totals not representative of individual sums

1. Nov 4, 2012

### liveitout

Hi there,
This is going to be difficult for me to explain, so I will try my best. My statistics is kind of rusty...
I've been given census data to analyze and I'm having problems. The totals are all given by the government.

For a specific attribute I've been given rows of different areas, and columns pertaining to the attribute. For ex. Race is divided into visible minority (further broken down into 5 other descriptive minorities) and non-minority.
The first row contains the total population, and total number of non-minorities, and each race and then the non-minorities. This all works out fine.

However, for each row (or area) the SUM total population is off by +/- 10.
Non-minorities + Visible minorities do not equal the total, and are either above or below by 10.
I'm trying to get percentages for each area, but if the total isn't accurate...what exactly can I do statistically to make this valid?

ex) Visible Minority = VM, non-minority = NM, Race1 = n, Race2 = k, race3= l

VM = n + k + l
VM + NM = total population
But in this case VM + NM = +/- 10 the given total

If you need further examples please let me know, I've tried my best!

Thanks a lot.

2. Nov 4, 2012

### Stephen Tashi

I suggest you give a specific numerical example of the data layout. If the real data is complicated, you might link to an example of it if it is public data and online. Or you might make up a simplified example.