OMB Winter Mtg.-February 5-6, 2015

103

LEAST COST FORMULATIONS, LTD. 824 Timberlake Drive, Virginia Beach, VA 23464-3239 Tel: (757) 467-0954 Fax: (757) 467-2947 E-mail: office@lcftld.com URL: http://lcfltd.com/

TECHNICAL REPORT

NUMBER:

TR322

DATE: 2012 July 8

TITLE:

Statistical analysis of interlaboratory studies. XXXV. Reproducibility estimates from proficiency test data.

AUTHOR: R. A. LaBudde ABSTRACT: The issue of estimating reproducibility effects from proficiency test (‘PT’) data is discussed. It is recommended that such data may be used to estimate

reproducibility when: 1) There are sufficient collaborators (8 or more remain in the final set of data); 2) Collaborator results are removed only for known cause based on subject-matter expertise and facts. It is recommended that the method used to estimate reproducibility effects is by the standard deviation of mean (across replicates, if any) results for the entire dataset net of crude errors with known causes, corrected for replication. If useful, a lower bound on reproducibility may be obtained conveniently via the interquartile range of the original dataset.

KEYWORDS:

1) PT

2) REPRODUCIBILITY 3) VARIANCE

4) ROBUST 5) OUTLIER

6) IQR

REL.DOC.: REVISED:

Copyright 2012 by Least Cost Formulations, Ltd. All Rights Reserved