Re: PCA C code contradictions

Reef Fish wrote:
Greg Heath wrote:
Reef Fish wrote:
Greg Heath wrote:
Reef Fish wrote:

Greg Heath wrote:
R1 = E[ x x'] is not used for PCA.
by Fukunaga is one) R1 is called the correlation matrix

RF > Fukunaga or Googamooga, R1 is NOT a correlation matrix.

Perhaps you MISSED the condition of the variables having
been standardized already.

No I didn't.

Otherwise, it's an ERROR, no matter in what discipline or in
what textbook it was written.

In your case, it was your misunderstanding between the covariance
matrix of standardized variables and unstandardized variables.

Just learn the basics and move on.

I thought that was the end of it, and didn't realize Greg came
back to make his last stab:

If you keep making mistakes, I have to keep coming back. The
solution is easy: Stop making mistakes.

Like the way you kept coming back on your misunderstand of the
role of simple correlations in multiple regression?

No. Like the way you kept coming back on your misstatements of
how I advised simple correlations to be used.

Every time you
come back, you merely add one more piece of evidence to your
ignorance on the subject.

But I soon found out that the college students in the USA were
practically illiterate in English - compared to what I learned in the
few years in high school.

All of the college students in the USA are practically illiterate?
That statement sure adds a lot of credibility to the rest of your
arguments.

That's no more hard to believe than the fact that you are nearly
statistically illiterate. It's hard for you to accept isn't it?

I accept truth. I reject your silly statement.

My knowledge of statistics has been pretty much limited to what
I had to know for my research in statistical pattern recognition.
In that realm I'm highly literate.

Just like you argued for weeks about how YOU can
get information about multiple regression from simple correlations
where nobody else, statisticians or non-statisticians can.

I see you've deleted my reply indicating that the above is a
misrepresentation of my claims.

And you made two mistakes in two different posts just writing down
the formula for the REAL correlation matrix! :-)

Excellent point! Four instances of typing V(-1/2) when I should have
written V^(-1/2) is inexcusable.

I stopped counting when it reached two. You should read the post:

http://groups.google.com/group/rec.travel.air/msg/d7ba035f284cb1d0?hl=en&;

"A Treatese on key MISSPELLINGS and inconsequential TYPOS".

Most typos are inconsequential, as the typo in the title above. But
when
someone (in that post) claims statistical expertise and then spells
several
times "statics" and "Stactics" for "statistics", they become key
misspelings
that betray the claim of the misspeller as an expert in statistics.

The same goes for your FOUR-TIME mistyping of the correlation
matrix formula,

There you go again. I mistyped the formula once. I cut and pasted it
another time. That's only TWICE. The formula contained V^(-1/2)
twice.

You need to slow down and get your 2s and 4s straight.

and then claim yourself to be knowlegeable about
correlations.


You are just a chronic excuse-making, know-nothing in statistics,
wasting everyone's time arguing the unarguable.

Give it a rest, and try to learn something form others, such as
Eric Bohlman in this case, if not from me.

Learn what? How to misrepresent the meanings of someone
else's posts? ... Spare me.

-- Bob.

.

Quantcast