Some notes:

I assume that you all are motivated not by grades but
by learning stuff. Hence, I have read your homeworks
carefully, and made comments, but have not scored them
in a particularly hardcore way. The scores are this,
for each problem: 2 -- good job, no significant problems;
1 -- some significant problem(s); 0 -- you didn't do it.

r measures how well the two data sets are correlated,
not how well the two data sets' errors are corrlated.

A probability of 99.95% is NOT approximately equal to
100% -- not when you are considering probabilities!

In your plots you MUST show the error bars. Otherwise,
how can you tell how good/bad your fits are? Without
error bars, your weighted LS fit looks a lot worse
than the unweighted fit.

How to do sigma when you have uncertainties in both
x and y? Several ways: (1) ignore sigma_x; or (2) take
whichever is bigger; or (3) use sigma_x*sigma_y instead
of sigma_y^2; or (4) combine the errors some other way
(but be careful -- if you just sum sigma_x^2 and 
sigma_y^2 then you are increasing the size of the errors --
did you mean to?).

What does a t test prove? That the *means* are the
same (or not the same) -- not that the *data sets*
are the same.

For the K-S test I wanted you to produce the sorted
version of the data, and plot it, and measure D -- not
to use an online tool. It's not all that hard to do.