On 27 Oct 2000 15:02:50 GMT, [EMAIL PROTECTED] (Dubinse) wrote:
I have the need to combine a number of data sets that are in the
form of histograms where the class marks and boundaries may vary
from data set to data set. The data represents numerical counts
as a function of particle size on an asummedly continuoous scale.
I want to combine some of the data to form a matrix with the rows
representing the individual data sets and the columns representing
the particle size on a single compatible scale. Eventually I want to
calculate the variance/covariance matrix etc. In some cases the
particle size scale is linear and in some cases it is logorithmic. I
have not seen such matters treated in the literature, but I am mostly
a country "hoss doctor" and not a professional statistician.
It seems to me that you might start with a set of box-and-whisker
plots. See Exploratory Data Analysis.
Questions:
(1) Are there methods described and software available for doing
such aggregation of data?
I did not see you mention something that I would consider
"aggregation." Correlation is something other.
(2) Are there pitfalls or violations of (sacred) assumptions involved
im making such aggregations that would render the
variance/covariance matrix invalid or otherwise debase further
statistical operations that might ensue?
How do you want to weight by Ns? Keep in mind that you are *not*
interested in testing, and that you want to keep your descriptions as
clean and neat as possible. And check, earliest, with a local
statistician.
(3) Is there some keyword, text, link etc I should examine in order
to dispell my ignorance.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=