A BIAS in the Pensieve: A Batty Idea About Snape

[nkafkafi]

Lyn now:
It can be fun to be "geeky and quantitative" sometimes but in this
case a t-test
is not appropriately applied and thus the results are not meaningful.
A frequency table
(which, of course is what Charme did) is appropriate, but the
assumptions for a t-test is
that the event can occur randomly. Now if it were the use of a word
like "other" that
might be used undeliberatively, you might consider its use at least
quasi-random and get by with it (and indeed I am sure you are aware
there are analysis routines that do just this).
However, words like hag and vampire are selected deliberately and thus
are no longer open to random inclusion in the text, thus a t-test
would be misapplied for this sort of
data.


Neri:
(Warning: "geeky and quantitative" stuff ahead)
Words like "other" are also used for some reason, not randomly.
Scientists believe *everything* happens because of some reason and
nothing is truly random (or at least, nothing above quantum level),
and yet they use t-test all the time. The question is not if it's
random, but if the reason JKR chose the word is indeed the reason that
we hypothesize. That is, suppose JKR describes the clientele of the
Hog's Head and she thinks: " I must throw in some non-human here to
make it more colorful. Now what will it be? A hag? A vampire? A
banshee? A veela? OK, lets have a hag because it just seems the sort
that would hang in the Hog's Head". In this case you'd expect that
overall vampires would be mentioned a similar number of times as other
non-humans. But suppose JKR goes like: "what non-human should I use?
Wait! I must put in some vampires and also supply some information
about them, so when I reveal in Book 7 that Snape is half vampire they
won't say I didn't play fair". In such a case you'd expect that
vampires would be mentioned considerably more than other non-humans
that don't play a key role in the story. This is the usual situation
in which t-test is used. Now, if you really want to catch me in a
statistical error you could have said something like "t-test assumes a
normal distribution and you don't know that. You should have used a
non-parametric test" which is perfectly true, only non-parametric
tests have less power than parametric tests, so I sincerely doubt that
a non-parametric test would have found a significant difference where
t-test didn't. 


Lyn:
You keep asking what point it would play in the plot, even though I
have given
Several examples of how it could, and both Charme and SSSusan have
mentioned the
Significance in their own views. To reiterate just one that SSS
discussed as well, it would
explain why Snape remained silent about the Prank, and why he may have
retained resentment over it.


Neri:
You can always suggest how to connect something to the plot. I guess
what Magda and I meant is to ask "what major mystery would it solve?"
I mean, Snape being a vampire or half vampire or whatever is a quite a
big bang. Bigger than Hagrid being half giant. Bigger even than Lupin
being a werewolf, which was a one-book mystery, not a five-book
mystery. It doesn't look like JKR would make Snape part vampire only
to plug a minor plot hole about the prank or something like that. It
should be something very important. I mean, if you said that he spies
on Voldy and the DEs by turning into a bat and hanging over their
heads, well I'm not sure I'd buy that, but at least it would be a
worthy reason to make Snape a vampire.

Neri