1400 New Englanders Determine Ratings In Boston TV Market

[Garrett Wollman]

<<On Sat, 28 Mar 2009 08:52:11 -0400, "Mark Watson" <markwats@comcast.net> said:

>   Found this interesting Boston Globe article about Nielsen's "People 
> Meters" that are in 600 households in the Boston TV market, 1400 people in 
> all determine ratings for a market of 2.4 million households. The article 
> questions whether the number of people meter families is too small and how 
> accurately what they watch are logged.

And of course the /Globe/ reporter completely missed the most
important statistical fact, which is that programming and advertising
decisions ("live or die by a tenth of a point") are being made on the
basis of random sampling noise.  The margin of sampling error for n =
600 is 4% (at 95% confidence).  They would need to sample many
thousands of households per market to bring that down to a tenth of a
point.

The standard error (another measure of sampling error) for n = 600 of
an 8.1% survey result is 1.1; this means that the true result,
nineteen times out of twenty, is between 5.9 and 10.3.

-GAWollman