• You are here: EPA Home
• CADDIS
• Stressor Identification
• Step 3: Evaluate Data from the Case: In-Depth Look
• Verified Predictions

Step 3: Evaluate Data from the Case

• Overview
• In-Depth Look
• Results and Next Steps

Verified Predictions

Concept

Knowledge of a cause's mode of action permits prediction and subsequent confirmation of previously unobserved effects.

Examples

Consider organophosphate pesticides as a candidate cause of reduced aquatic insect abundance. What findings support or weaken the case for these pesticides as the cause, based on evidence garnered from predicted effects?

• Supporting evidence - Monitoring data show that other taxa sensitive to cholinesterase inhibitors (e.g., crayfish) have declined at the impaired site, while taxa without cholinergic receptors (e.g., rotifers) have not been adversely affected.
• Weakening evidence - Monitoring data show that taxa predicted to decrease with exposure to organophosphates are not reduced at the impaired site, whereas taxa predicted to be unaffected have declined.

How do I Analyze My Data?

Prediction requires knowledge of effects that occur with particular causes. If other candidates may cause the effect, successful prediction of the effect is only weakly supportive. The prediction may involve evidence of temporal sequence, spatial/temporal co-occurrence, stressor-response relationships, exposure or mechanism, causal pathway, or manipulations. For this reason, special care should be taken to avoid biasing the analysis by using the same evidence more than once. This type of evidence can greatly strengthen the case for a candidate cause, if it truly results from a prediction, because predictions are based on a deeper knowledge of mechanisms driving potential cause-effect interactions.

Figure 3-10. Verified Predictions. A clever assessor predicts that if the impairment (dead fish) is due to a cholinesterase inhibiting pesticide, he would find that crustaceans would also be dead, but organisms without cholinergic systems would be alive. If he then searches for evidence at the site and finds dead crayfish and live rotifers (upper right panel), that strengthens the candidate cause. If the crayfish are alive (lower right panel), that weakens the candidate cause.   [general explanation of illustration]

The logical constraints and statistical tests appropriate for analysis of predictions depend on the nature of the prediction or the type of evidence the prediction addresses; these issues have already been discussed under previous types of evidence from the case.

Examples of analysis in support of Verified Predictions:

• Using predictions of environmental conditions from biological observations
• Using species traits

What evidence would support or weaken the case for a candidate cause?

Supports

• Data showing that a predicted, previously undocumented effect is observed at the impaired site

Weakens

• Data showing that a predicted, previously undocumented effect is not observed at the impaired site

Refutes

• Data showing, with a high degree of certainty, that a predicted obligatory effect does not occur at the impaired site

How do I score the evidence?

Helpful tips

• The power of this type of evidence resides in making predictions, rather than explaining the reasons for observations post hoc; observation without a prior prediction is simply descriptive.
• The best predictions are specific and unambiguous.

Top of page

Area Navigation

• CADDIS Home

---

• Stressor Identification
  • Introduction
  • Step 1: Define the Case
  • Step 2: Candidate Causes
  • Step 3: Data from the Case
  • Step 4: Data from Elsewhere
  • Step 5: Identify Probable Causes
  • Types of Evidence
  • Scores
  • Causal Assessment Background

  ---

• Sources, Stressors & Responses

---

• Examples & Applications

---

• Data Analysis

---

• Causal Databases