National Assessment of Pulp and Paper Environmental Effects Monitoring Data: Findings from Cycles 1 through 3
- Publishing Information
- Acknowledgement
- Abstract
- 1.0 Executive Summary
- 2.0 Introduction
- 3.0 Overview of Studies Conducted in Cycle 3
- 4.0 General Methods - Data Preparation and Analysis
- 4.1 General Methods - Procedure for Determining National Response
- 5.0 Fish Survey
- 5.1 Data Processing and Study Designs
- 5.2 Summary of Effect Sizes
- 5.3 Response Patterns and Meta-analyses
- 6.0 Fisheries Resources and Usability
- 7.0 Benthic Invertebrate Community Survey
- 7.1 Data Processing and Study Designs
- 7.2 Summary of Effect Sizes
- 7.3 Response Patterns and Meta-analyses
- 8.0 Sublethal Toxicity Testing - Introduction
- 8.1 Sublethal Toxicity Testing - Monitoring Changes in Effluent Quality Among Cycles
- 8.2 Sublethal Toxicity Testing - Summary and Future Considerations
- 9.0 Summary and Conclusions
- Glossary
- Acronyms / Abbreviations
- References
4.1 General Methods - Procedure for Determining National Response
Meta-analysis is a set of statistical procedures used to quantitatively synthesize the results of a large number of independent studies. Further, it permits overall response patterns to be determined. The meta-analyses required determination of a standardized magnitude of effect. For the more commonly employed control/impact designs, the Hedges’d effect size was calculated as the difference between the exposure and reference means, divided by the pooled standard deviation (this value is multiplied by a correction factor that accounts for the effects of small sample sizes) (Rosenberg et al. 2000). When pooling control/impact and gradient design data, it was necessary to use a second, related effect size measure, the Fisher’s z-transform (Rosenthal 1991, Rosenberg et al. 2000, Lowell et al. 2003).
The main meta-analytical results are presented in the following summary format (Fig. 1). The standardized effect size is on the x-axis, with the vertical line representing a zero effect. The result for each mill grouping is presented as a horizontal 95% confidence interval about a vertical tick mark indicating the average effect size for that grouping of mills. Mill distributions to the right of the zero effect line indicate that the average effect associated with effluent exposure was an increase in the measured endpoint. Similarly, mill distributions to the left of the zero effect line indicate an effluent-associated decrease in the measured endpoint. The increase or decrease is statistically significant for the group as a whole if the 95% confidence interval does not overlap the zero effect line. Larger mill groupings (that are non-significant as a whole) can be composed of smaller subgroups, some or all of which may be significantly different from zero. All of the meta-analysis results in the following sections will use this graphical representation of the data.
Figure 1: Example of a meta-analysis summary figure. The effect size was measured as either Hedges’ d or Fisher’s z-transform (see text).
- Date Modified: