"Brenda Ann" wrote in message
...

"David Eduardo" wrote in message
...
In each market area, all listening to any radio station is recorded by
listeners as is the instruction in the Arbitron diary. Commercial or
non-commercial, local or not, internet or off air, satellite or
terrestrial. All is recorded and processed. If there is any significant
listening to out of market stations it is recorded.



WRONG. Arbitron does NOT log "ALL LISTENING". They log a small percentage
of listening, and profess to know what all the rest are doing based upon
that.

You know what I meant: the Arbiytron diarykeeper logs all listening and all
that listening is processed.

Statistics are crap. They are not, and really can never be, accurate.

They are accurate within a margin of error that is easily calculated, and is
small enough for advertisers to spend $21 billion on radio this year.

Statistics is a science, and it has the unique quality that "error" is not a
dirty word. Speaking of samples, when you last had a blood test, did they
take all your blood, or just a small percentage? They took a sample, as they
know that it would faithfully represent all the rest of your blood. This is
exactly what a good poll does; a good sample can be tested, as Arbitron has
done, by a replication study where the same thing is done twice to see if
the reuslts are the same... and they are.

They are mathematical sleight of hand. Smoke and mirrors. I doubt there is
any real scientific foundation for them at all, since it's highly unlikely
that anyone did a small sample, then went to six million people and asked
each of them the same questions to verify the numbers.

That is not how you test a poll. It is done by a replication study. You do a
sample, then repeat it. If you get identical results, the sample size and
procedure is valid.