NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/ai/neuraln/2951

Text File | 1992-07-25 | 1.6 KB | 37 lines

Newsgroups: comp.ai.neural-nets
Path: sparky!uunet!wupost!zaphod.mps.ohio-state.edu!menudo.uh.edu!lobster!nuchat!texhrc!texhrc!ak45ldp
From: [e-mail] (Larry D. Pyeatt)
Subject: Re: SImple pattern recognition problem - any answers?
Message-ID: <[e-mail]>
Sender: [e-mail]
Nntp-Posting-Host: aisun
Organization: Texaco
References:  <[e-mail]>
Date: Fri, 24 Jul 1992 19:31:53 GMT
Lines: 24

In article <[e-mail]>, [e-mail] (Alan Wilson) writes:

In most cases, I would say that 5 examples of each class is not enough
to prove anything.  However, your data MAY indicate that by knowing a
person's scores on the test you may be able to deduce his classification
as A or B with a high degree of confidence.

Alternately, you may be able to deduce a person's likely responses to the test
questions by knowing whether he is an A or B.  The data that you have seems
to indicate that the classes are nearly linearly separable.  If that is
the case, it would be easy to build a pattern classifier to tell whether
a person is A or B.

Also, you could possible learn to estimate a persons PC2 (or PC1) score
by knowing their PC1 (or PC2) score and whether they are an A or B.

However, I do not think that you have enough data to draw any strong
conclusion.  Get MORE data and TEST for statistical significance!

--
Larry D. Pyeatt                 The views expressed here are not
Internet : [e-mail]    those of my employer or of anyone
Voice    : [phone]       that I know of with the possible
                                exception of myself.