Jym Dyer wrote:
Frank Krygowski writes:
So what you're saying is "_All_ the data is bad. I
just know it is. You just have to believe me."

=v= What I actually said is that *part* of the data is
subject to bias: namely, the at-fault numbers.

And again: Where is your evidence?


Sorry. I think unquestioning faith in a self-proclaimed
prophet is kind of dangerous.

=v= Now you're really off the rails. I named specific and
well-known forms of bias (observer bias and survivor bias).
I did not invent these insights, I'm just applying them.

But you haven't shown that it applies to the Phoenix data, or if it
does, to what degree. You can hardly expect researchers to discard all
data just because Jym Dyer thinks it might be invalid!

If you have any substantial, non-_ad_hominem_ reason why these
biases should simply be ignored, by all means present them.

Until they're demonstrated to exist, they must be ignored.

Understand, I'm not claiming all Phoenix data is faultless. I'm aware
of problems with data collection. In fact, I've corresponded
extensively with two statisticians who are recognized masters at
spotting shortcomings and bias in similar data collection.

But those individuals don't just toss all data based on the idea it may
contain bias. They are known for finding evidence of such bias, when
applicable, and using their evidence to improve the analysis of the data.

So far, you've done none of that. You've just said "Ignore that data."
So seriously, what have you got for evidence that Phoenix data is
significantly in error?


--
- Frank Krygowski