SYMPOSIA PAPER Published: 01 January 1953
STP46776S

The Importance of Correlations in Detergency Testing

Source

About three years ago, attempts were made in our laboratory to apply statistical methods of correlation analysis to a series of comparative detergency tests. Some of this work was later published by Lambert and Sanders (10), but none of the results obtained by quantitative correlation analyses needed inclusion; this was fortunate, since the inadequacy of the existing statistical methods for determining the reality of an apparent relation between two testing methods has meanwhile become obvious. The difficulties are particularly serious if only six or eight detergents are used for providing the experimental data as in this case. Recently, Harris and Brown (7) published two papers whose main topic was the comparison of wash test methods and of soiled test fabrics. With the exception of one set of data in the second paper, only four detergents were used in the various tests; nevertheless, the correlation coefficients based on such sparse information were tested for significance and the tests ranked accordingly. As will be shown later, this type of treatment, making little allowance for the fact that the calculated coefficients are only very doubtful estimates of the true values, passes up truly high correlations and mistakes negligibly small correlations for important ones. There is no doubt that correlations should be studied extensively in detergency testing work. As one of the present authors has pointed out previously (9), the prime objective of laboratory procedures for the evaluation of detergents is to rank commercial or experimental products in accordance with their performance in practice. Therefore, it is apparent that the ultimate evaluation of the usefulness of a particular test method depends on a careful comparison of the results obtained with those available from end-use tests. Furthermore, in the development of improved laboratory methods, it is frequently of importance to determine the degree of concordance between various procedures. A case in point is a comparative study, which was made in our laboratory, testing ten detergents in an abridged multicycle test and in a single cycle test with an experimental soiled fabric. While preliminary results of the correlation analysis were given elsewhere (9), it was thought of interest to reinvestigate the available statistical methods more thoroughly and to determine just how much useful information could be extracted from the data. Before presenting the various steps of the statistical correlation analysis, it seems appropriate to discuss some of the simpler methods for studying correlation or lack of correlation qualitatively or semi-quantitatively. The usefulness of these latter methods will also be illustrated by a few examples taken from practice.

Author Information

Lambert, Joseph, M.
General Aniline and Film Corp., Easton, Pa.
Leshan, Edward, J.
General Aniline and Film Corp., Easton, Pa. California Research and Development Co., Livermore, Calif.
Price: $25.00
Contact Sales
Related
Reprints and Permissions
Reprints and copyright permissions can be requested through the
Copyright Clearance Center
Details
Developed by Committee: D12
Pages: 27–38
DOI: 10.1520/STP46776S
ISBN-EB: 978-0-8031-6960-9
ISBN-13: 978-0-8031-6522-9