Hi Bill,
If you check out this link:
http://en.wikipedia.org/wiki/Sampling_method,
you can see that it talks extensively about developing a problem that needs
quantification, developing a sampling method and then a sample frame. This
article discusses sample sizes as well as data sources within a given area.
My problem with the provided numbers from Arne are that, number one, he only
sample one type of source (online), and has deemed that type of data to be
the only type needed to confirm his result. He has ingnored the many
non-online data sources. Tom, then went on to change the problem to be
quantified in mid-stream from just VB.NET vs. C# jobs to Senior Developer
jobs and then dismissed non-online sources althogether.
So, for both Arne and Tom, they have limited their source data to just one
of many possible types.
I then argued that Arne's limited sample type is, in and of itself, too
small to provide an accurate sampling for all online sources (which the link
provided above backs up in the sample size section).
From the Brown paper, pages 4-10 disprove Arn'e *probability problem* and
his supposition that 3:1 is "statistical evidence", rather than
"observational" or "anecdotal" evidence at best.
Because Arne only looked at one of the many possible sources of data, page 2
in the paper agrees with my supposition that he is not working with
statistical *evidence*.
-Scott