Thursday, May 10, 2012

Statistics are a like a Loaded Gun

And frankly some people should not play with them unless they know what they are doing.

Have a look at this christian blog's post.

He posts the graph of PZ's blog hits on Science Blogs,
There is a drop off in hits, a significant one, in fact.  So what does he do?  Look for reasons? Investigate the data?  Look for possible reasons.
Not even close.  He claims victory and says god is on his side.

ah. OK.

Of course there is the little matter that PZ moved over to Freethought Blogs in August, the same time the drop off occurred.

PZ goes into this into more detail (and more for his own enjoyment) that I do.

Here is the Hard Truth.
If you are going to draw conclusions based on a one off observation you better have the data to back it up.  If you are going to make a conclusion based on data, then you better have the hypothesis to be able to predict the data.

See here is how we really do science.
Statement: Christian's are "winning" (whatever that means)
Hypothesis: We should see a difference in blog hits at PZ's blog.
H0: Hits are the same across the last year or greater.
H1: Hits are less.

At this point Pastor Tom looks at the data and runs around screaming victory.
The trouble is that the hits are not down.  If anything they are up in total.
So we have to reject H1.

Science. It just works.

Sorry Pastor Tom.   But your post proves nothing at all.