The mathematical study of test plot data

IRG/WP 2282

W H Hartford

It has long been desirable to have a mathematical expression, which, by the insertion of statistically derived constants, would describe the behavior of test specimens in service. The philosophy of such expressions is briefly discussed, and the history of the concept is outlined. The modern approach to the problem dates from the graphical presentation of dosage-response and deterioration curves by Colley (1970). In 1972 Hartford devised the "Performance Index" giving a numerical parameter which could give statistically valid comparisons between preservatives in a given installation. A more potent tool is the "log-probability" method, which is now under consideration by AWPA P-6 as a standard. If "log score" is converted to probits x the relationship: x = a + b ln t + c ln R holds well for penta and creosote test specimens at log scores above 30, and tends to give conservative estimates of performance when early results are evaluated. With CCA, this is not true. Recent availability of 19-20 year test results on CCA stake and panel tests (Leach 1986) has permitted the evaluation of "modified probits" as the x in the above equation. Preliminary work shows promise: the CCA results are improved and low log scores can be used. Preliminary data are available on various CCA formulations and stake vs. panel tests.


Keywords: FIELD TEST; LOG-PROBABILITY METHODS; PANEL TESTS; STAKES; CCA; PCP; CREOSOTE; TEST SITES

Conference: 87-05-17/22 Honey Harbour, Ontario, Canada


Download document (536 kb)
free for the members of IRG. Available if purchased.

Purchase this document