[thelist] flash detection

[deke]

On 7 May 2001, at 17:18, John Dowdell posted a message which said:
> (Deke, check NPD's methodology papers... their Online Panels are
> representative consumer samples with a 95% confidence level. No
> sub-sampling is ever perfect, true, but this one seems pretty good.)

Representative of *what*?

Is there a difference between someone living in the Over The Rhine
neighborhood of Cincinnati and someone living 30 miles from the 
nearest Pepsi machine in rural Montana?  No. They are all "central 
US".

Is there a difference between a single retired homeowner without
a mortgage, and a taxfree income of $24,500 and someone trying
to raise five kids by working full time at $12,000/year before taxes?
No. They are both "under $25K"

Is there a difference between an 18-year-old who barely knows how
to say "Jew won frizwit dat?" and a 55-year-old master machinist?  
No. They are both high school graduates.

There is *nothing* in the NPD methodology that suggests that their
panel members are representative of others in terms of their hardware
or in terms of their internet access, and the fact that they are completing
surveys on the remote chance that they *might* win a token prize
suggests that these are people who do *not* value their time highly.

There are better ways to select a panel. You might take a look at
how InterSurvey (part of KnowledgeNetworks) chose the people they
wanted for their panels. KnowledgeNetworks was started by the
same guy who runs SPSS, the stats software that's been dominant 
in sociological studies for more than a quarter century.

deke





------------------------
 "The church is near but the road is icy; 
  the bar is far away but I will walk carefully." 
                            -- Russian Proverb