Perhaps I misunderstand your intent, but I have the impression that you
are seeking a way to deal with your data that doesn't require a lot of
hard thinking about the meaning of your data and the foibles of various
analytical methods.   If that is indeed the case, my only comment is
TINSTAAFL (= "There Is No Such Thing As A Free Lunch").

Given the information so far available, the only advice I have is to
display the data as plainly as possible.  This may mean in tables (not
necessarily restricted to two dimensions), in scatterplots, or in other
ways.

 It is difficult to know what useful advice to offer, not least because it
is quite unclear what your variables are.  You speak of "activities", and
we (that is, those who have responded to your queries) have assumed that
these refer to categorical variables having only two values:  on/off, or
present/absent, that may reasonably be coded (1/0) or (1/-1).  But you
also refer to "sensors", and you have not said either (1) what the sensors
are sensing nor (2) what kind of information they report (again, at least
one respondent has assumed that these variables also are binary (on/off),
but in my lexicon a "sensor" may equally well report an intensity (volume
or loudness on some scale), a position (in one, two, three, or four
dimensions (one dimension of time and up to three of space), a quantity
(amount of fuel remaining, operating temperature, ...), a count (number of
items satisfying a particular criterion or category);  inter alia).

And for that matter, you have not even specified whether your "activities"
are mutually exclusive, either in whole or in part:  information relevant
to how one might imagine making a display of elementary information.

While you have referred to "regression analysis" as a style of analysis
you would like (one gathers!) to invoke, you have not said which of your
variables you conceive of as possible predictors, and which as possible
response variables.

For completeness, I include below the three postings (labelled [1], [2],
and [3]) I have seen from you.  If you believe you have indeed told us
information that I have described above as lacking from your input, I'd
be interested to know where you perceive it.

 ------------------------------------------------------------------------
Date: 30 May 2003 08:14:10 -0700                                [1]
From: Fahd <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Subject: Regression Question

  Does anyone know how would it be possible to convert discrete data to
continous data for regression analysis?  What are the implications of
doing that?  Also, is there any good tool for plotting regression
equations (logistic curves etc.).  Thanks in advance.


Date: Fri, 30 May 2003 09:30:27 -0700                           [2]

Sorry for not being clear on that.  The problem is very simple.  We have a
room with all sort of sensors (light, weight, software sensors...etc) and
we have records of the activities people are doing along with readings
from the sensors.  We want to identify the correlations between the
sensors and the activities, for example, if it is say a presentation then
the projector sensor should show a high correlation with the activity.
We are of course representing the activities with discrete values. Any
thoughts about the implications of using regression.


Date: Sat, 31 May 2003 09:55:36 -0700                           [3]

Thanks a lot for the info, but I have one other simple question.
Encoding activities as 0 or 1 will allow me to identify the sensor
correlations for these two activities.  But what if I have a third
activity, is there a way to figure out the set of sensors that correlate
to the three activities instead of looking at them in pairs or is there a
way to combine the correlations from the pairs?  Thanks.

 -----------------------------------------------------------------------
 Donald F. Burrill                                         [EMAIL PROTECTED]
 56 Sebbins Pond Drive, Bedford, NH 03110                 (603) 626-0816

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to