Statistics (was `Out of India')

[Vidhyanath Rao]

[This thread may not be of any direct interest to most members of
Indology. I am posting this publically because I wish to
correct a potential misunderstanding which my previous post may
have created. I would suggest all purely statistical discussions be
switched to private e-mail.]

Kishore Krshna <kishore at mail.utexas.edu> wrote:

>How many pairs of observations did you have for the analysis? If it
>was less than 30 (the central limit theorem threshold) underlying
>assumptions about normality may not hold [...]

Rank correlation is supposed to be a non-parametric test, so the
underlying distribution should not affect the distribution of the
statistic.

Kolmogrov-Smirnov test is supposedly for different samples drawn from
*continuous* populations. I am not sure why it would be a better test
for comparing two rank orders. I am not even sure how I would apply
since the `observations' are just the ranks from 1 to 9 in both
orders.

>[...] However, what is the theory that suggests that you 
>should be examining these specific entities? Why these specific works and
>those specific parvans? Wouldn't it make more sense to do this across
>all parvans of the Mahabharata to test opinions about interpolations? [...]

The point of the experiment is to test the claim that differences
in meter indicate differences in age or authorship. What you are saying
makes sense if I had already accepted the method as valid and then
try to use it to look for interpolations in Mahabharata etc.

The general idea is that when I start using something new, I try it
out in known situations first. If it works there, then I can think
about using it in situations where the answer is not known.

In other words, I was not testing for interpolations in Mahabharata or
Kumarasambhava  etc., but I was testing the idea that statistical
differences in metrical patterns of anushtub (that is half-stanzas of
16 syllables, ending in LHLX, L=light, H=heavy, X=any) can indicate
differences in age or authorship.

-Nath
Nath Rao (nathrao+ at osu.edu)		614-366-9341