Separation Point Analysis Method (SPAM)

PDF

Authors
  1. Hunter, D.G.
  2. Edmond, E.J.
Corporate Authors
Operational Research Div, Ottawa ONT (CAN) Central Operational Research Team
Abstract
This research note documents an extension of the iterative post-hoc analysis method described by Emond and Massel (2003) to positive integer-value datasets. The original assumption was that the data were drawn from one or more normal distributions. An additional data model has now been added in which the data are assumed to be derived from one or more binomial distributions. Using an appropriate test statistic and likelihood function, this method finds the most likely separation points between datasets that have been shown to have binomial parameters that are statistically significantly different. The method is of particular interest within operational research due to three important properties: it always returns an answer, this answer is never intransitive, and its error accumulation is multiplicative rather than additive. Both the binomial data model and the normal data model of Emond and Massel have been implemented in a MATLAB software application known as SPAM (Separation Point Analysis Method).

Il y a un résumé en français ici.

Keywords
Multiple comparisons;statistical testing;post-hoc analysis
Report Number
ORD/CORT-RN-2004-11 — Research Note
Date of publication
01 Jul 2004
Number of Pages
47
DSTKIM No
CA025685
CANDIS No
523403
Format(s):
Hardcopy

Permanent link

Document 1 of 1

Date modified: