Originally Posted by Kosty
This is different as it is not a random sample. It is very stratified and the various retailers large and small do not count the same.
Political polling or normal statistical numbers do not apply for a non-random sample.
The Nielsen videoscan numbers in this case is no longer a statistical exercise with random sampling. If I understand their methodology, they have retailers responsible for 40% DVD sales participating, and their data is collected at the register as the sales are rung up. Barring any mechanical errors or failure to submit data, their sales numbers for participating retailers are complete. And this percentage of the retailer population that participates is high.
IE: unlike statistical sampling which deals with a small percentage of the population, what they have is complete for the participating group. This size is beyond any normal statistical sampling considerations, so it is not correct to consider this the same as other nielsen ratings for TV shows (which is really about tiny statistics sampling sizes with a low percentage of the population participating, and yet is trusted by practically anyone. Nielsen ratings to TV shows are used to decide which shows live or die and rates for commercials even when they are much less complete.
As has been pointed out various times by many individuals here, amazon participates in nielsen, which should make tomorrow's HMM release interesting as the April 15th group buy will be part of that. We should see how much the nielsen numbers can be skewed by the impressive amazon showing of the red crusaders.
--- IMO, this thread gets interesting only on Thursdays and Fridays, the other days, this could just become a chatline for lonely people.