Data Files

SERIES B
Statistical Methodology

Bayes model averaging with selection of regressors
P. J. Brown, M. Vannucci and T. Fearn
J. R. Statist. Soc. B, 64 (2002), 519 - 536

The sugars data used in this paper are described in
Brown, P. J. (1993) Measurement, Regression and Calibration.
Oxford University Press.

The data was provided by Shell Ltd at Sittingbourne.

The training or learning data consist of 125 records of near-infrared (NIR)
absorbance spectra (2nd differenced). There are 700 absobances
in each record, from 1102 to 2500 nanometers (nm) in steps of 2nm.
Each record begins with the sequence number (1-125).
The data are in file nsugar.dat.
Each record corresponds to a mixture of three sugars: sucrose, glucose
and fructose in aqueous solution, each at 5 levels (6,10,12,14,18 percent
by mass) in a full
3^5=125 design. These mixtures are given in
file nsugar.res (in order but without a sequence identifier).

The validation sample of NIR spectra consists of 21 records in file
nsugpred.dat. These correspond to the mixtures in an incomplete
'outer' design at 3 levels (0, 12, 25 percent by mass) in
file nsugpred.res.

It is desired to predict composition from spectra. In the paper this
is accomplised by performing a multivariate linear regression of
composition on spectra.
Software is available at http://stat.tamu.edu/~mvannucci/webpages/codes.html
Other relevant papers can be accessed from
http://www.ukc.ac.uk/ims/

P. J. Brown
Institute of Mathematics and Statistics
Cornwallis Building
University of Kent at Canterbury
Canterbury
Kent
CT2 7NF
UK

E-mail: Philip.J.Brown@ukc.ac.uk

• Dataset (dataset.zip, size - 411kb)
  

SERIES A
Statistics in Society

SERIES B
Statistical Methodology

SERIES C
Applied Statistics

SERIES D
The Statistician