Non-linear Regression of Preservative Performance Using Multivariate Probit or Weibnll Distribution Models: Extensions of the Classical Hartford and Colley Approach
The performance of wood preservatives is typically reported in the form of depreciation curves, which depict the average decay rating vs. exposure time for different preservative retentions. This method of data reporting has remained essentially unchanged for the last 65 years, despite researcher’s efforts to introduce more statistically founded methodologies for characterizing performance.In their seminal 1984 paper: “The Rationale of Preservative Evaluation by Field Testing and Mathematical Modeling,” Hartford and Colley presented a powerful log-probability method for wood durability analysis, based on a probit (normal fit) transformation of the data. Despite its ease of implementation and the wealth of information that is obtainable, the Hartford/Colley approach is still widely unknown and tittle practiced by the wood preservative community today.In this article, we review the steps required to perform such non-linear regression analyses of wood durability data using Microsoft Excel. This is illustrated using Lonza internal CCA data from a 1990 study featuring an emulsified oil additive to enhance pole climb-ability.The Hartford/Colley approach is expanded to include multiple covariates, enabling a reanalysis of Fahlstrom and % inch field stake data sets for different Cu-based preservatives exposed in Gainesville, FL. Instead of regressing each discrete preservative system and time against the average rate, our multivariate approach uses the individual components within preservatives as variables with all candidates analyzed together. Finally, the Hartford/Colley approach is extended to consider flexural decay in an AWPA E14 soft-rot laboratory soil bed test, comparing the efficacies of amine- and micronized copper with various co-biocides.
展开▼
