CADDIS Volume 1: Stressor Identification
Step 3: Evaluate Data from the Case
On this page
- Assembling the data
- Analyzing associations
- Types of evidence that use data from the case
- Using the evidence to evaluate candidate causes
Types of evidence
Causal analyses often begin with an examination of data from the case at hand. For example, a field biologist might observe that effects occur when a particular candidate cause is present, but do not occur when it is absent. Such associations provide the core of information used for characterizing causes. We recommend that associations from the case be evaluated first, because they can be powerful enough to eliminate candidate causes from further consideration.
Associations derived from other cases or biological knowledge cannot be used to refute a candidate cause, but can provide useful supplemental information for comparing strength of evidence. For example, a common assessment method compares ambient chemical concentrations with concentrations causing effects in laboratory studies; however, this type of evidence has substantial uncertainties associated with the type of laboratory organisms, extrapolation from lab to field, and so on. This and other types of evidence that bring in data from outside of the case are described in Step 4.
In Step 1, the biological impairment was defined and measurements were assembled that could be used to generate evidence to support or weaken a causal linkage in the causal pathway. In Step 2, candidate causes are selected for the analysis occurring in Steps 3 and 4. In Step 3, the evidence from the case is developed. The strongest type of evidence either supports the relationship between a candidate cause (a proximate stressor) and the biological impairment. However, other parts of the causal pathway also can be analyzed and provide evidence. More detail is provided in Assembling Data, which discusses data from the case, and Organizing Data along Conceptual Pathways, which lists the types of measurements that might be used to develop evidence.
We recommend developing a table that clearly shows the measurements that are available, and how they relate to each candidate cause (example tables of measurements). For evaluating uncertainty and data quality, additional tables should show the number and type of samples and provide references for the methods.
The SI process does not require a minimum data set, and existing data often are sufficient to determine the cause of impairment. However, you have the responsibility of evaluating whether the data used are of sufficient quantity and quality to support the causal analysis. You may suspect that available data are too sparse to support a causal analysis; however, it still may be useful to go through the SI process at a screening level, with the objective of identifying the data that would be most fruitful to collect. If you decide to generate additional data, what to sample and sampling designs are critical for success (Assembling Data) and the quality must be ensured (Data Quality).
Data are analyzed in terms of associations that might support, weaken or refute a candidate cause, using the types of evidence discussed below.
Volume 4: Data Analysis provides a section that helps you prepare and organize your data before analysis, and to evaluate data sources and quality. This volume also provides suggestions for selecting an analysis approach for your data. You also may download software which describes methods you can use to analyze your data. Some examples of analyses you can run with this software include scatter plots, correlation analysis, box plots, regression analysis, conditional probability analysis, and predicting environmental conditions from biological observations.
If you have listed multiple stressors as a candidate cause, the analyses of data from the case should be based on those aggregate causes. For example, if all divalent metals or all polycyclic aromatic hydrocarbons have been combined using a concentration additivity model, then all analyses of associations of that cause with effects should be performed using the sums of toxic units rather than the individual concentrations.
Additional specific issues are discussed in each type of evidence's information page (see links in Table 3.1. below and in the navigation bar in the upper right).
The evidence generated by analyzing associations among data or observations from the case will fall into one of the types listed in Table 3.1. It is important to avoid double counting by not using a piece of evidence as more than one type.
|Type of Evidence||The Concept|
|Spatial/Temporal Co-occurrence||The biological effect is observed where and when the causal agent is observed and is not observed in the absence of the agent.|
|Evidence of Exposure or Biological Mechanism||Measurements of the biota show that relevant exposure has occurred or that other biological processes linking the causal agent with the effect have occurred.|
|Causal Pathway||Precursors of a causal agent (components of the causal pathway) provide supplementary or surrogate evidence that the biological effect and causal agent are likely to have co-occurred.|
|Stressor-Response Relationships from the Field||The intensity or frequency of biological effects at the site increases with increasing levels of exposure to the causal agent or decrease with decreasing levels.|
|Manipulation of Exposure||Field experiments or management actions that decrease or increase exposure to a causal agent decrease or increase the biological effect.|
|Laboratory Tests of Site Media||Laboratory tests of site media can provide evidence of toxicity, and Toxicity Identification Evaluation (TIE) methods can provide evidence of specific toxic chemicals, chemical classes, or non-chemical agents.|
|Temporal Sequence||The cause must precede the biological effect.|
|Verified Predictions||Knowledge of the causal agent's mode of action permits prediction of unobserved effects that can be subsequently confirmed.|
|Symptoms||Biological measurements (often at lower levels of biological organization than the effect) can be characteristic of one or a few specific causal agents. A set of symptoms may be diagnostic of a particular cause if they are unique to that cause.|
The associations are evaluated by considering the degree to which they support or weaken the case for a candidate cause. We recommend scoring the evidence using a standard set of scores. These scores are described on each type of evidence's information page and are compiled in the summary table of scores.
Evidence based on case-specific data can be strong enough to eliminate an improbable cause from further consideration. The objective of elimination is to logically establish that a candidate cause is extremely improbable, or better yet, impossible and could not have produced the effect of concern. Evidence with strongly negative scores may be sufficient to refute, and thus eliminate, the cause. Because eliminated causes are not evaluated further, you must have sufficiently high confidence in the evidence to support this decision. When in doubt, the cause should be retained.
Confidence in the reliability of symptoms may be high enough to diagnose a cause. Similarly, the absence of symptoms that are always associated with a cause may be compelling enough to eliminate the cause from further consideration.