Designs to Minimize Loss of Information in Polynomial Regression

G. V. Dyke

Journal of the Royal Statistical Society Series C, 1974, vol. 23, issue 3, 295-299

Abstract: In reporting an experiment that includes a factor applied at many levels one may be willing to neglect certain terms of an assumed polynomial response curve. E.g. if there are 7 levels one may assume a cubic curve and neglect the quartic, quintic and sextic terms. If the block size is less than a replicate a “best” confounding can be selected by use of a standard computer program. “Best” here means “with minimum loss of information on (say) linear, quadratic and cubic terms”. The analysis of such an experiment involves covariance on a few simple dummy variates. The procedure for design and analysis for 7 or 8 levels in blocks of 4 is given in detail and the method of extension to other cases is indicated. The paper is presented as a modest example of the possible use of electronic computers to scan many possible designs and to indicate which of them are best using simple criteria, but without making exact orthogonality a sine qua non.

Date: 1974
References: Add references at CitEc
Citations:

Downloads: (external link)
https://doi.org/10.2307/2347122

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:bla:jorssc:v:23:y:1974:i:3:p:295-299

Ordering information: This journal article can be ordered from
http://ordering.onli ... 1111/(ISSN)1467-9876

Access Statistics for this article

Journal of the Royal Statistical Society Series C is currently edited by R. Chandler and P. W. F. Smith

More articles in Journal of the Royal Statistical Society Series C from Royal Statistical Society Contact information at EDIRC.
Bibliographic data for series maintained by Wiley Content Delivery ().