Political Conventions' Neilsen Ratings

Digby is puzzled by the Neilsen ratings for the convention speeches by Obama and McCain.

Neilsen numbers can be found here: McCain, Obama

My comments (a version of these also posted on Digby's blog):

From the link provided, Neilsen says, McCain's speech:

All Households (rating) 24.6
Number of Households = 28,298,000
Persons 2+ (rating) 13.4
Number of people = 38,933,000

Continuing there to find Obama's numbers:
All Households (rating) 24.5
Number of Households = 27,716,000
Persons 2+ (rating) 13.4
Number of people = 38,379,000

First, the difference in the first number (all households rating) is in the third decimal place (24.6 versus 24.5). Since a sample of 1000 typically has an error of 3%, to trust the third decimal place (1 part in a thousand, or 30 times smaller error) the sample has to be 30*30 = 900 times larger. i.e., the Neilsen sample cannot be that accurate unless the sample size is around **one million**.

Since I doubt Neilsen has a sample size that large, the correct answer is that the audience sizes are a statistical dead heat.

Please note the second number, the 2+ persons rating is identical - 13.4 - but the number of people is different. So, in the 2+ people households, if McCain reached more people in the identical number of households, it must mean that the McCain-watching households are slightly larger. So more kids watched McCain Big deal.

The correct answer again is really that it is a statistical dead-heat.

PPS: deleted part of the post. The persons/Neilsen rating point is not consistent between Obama and McCain; e.g., McCain's persons/hispanic rating point is 4% higher than Obama's and Obama's persons/white rating point is 1% higher than McCain's. But add the individual numbers and they don't match the totals, so Neilsen is applying some kind of a correction factor.