Conclusion Sensitivity Profile

Knowing that your conclusion is significant is not the same as knowing whether it is stable. The Conclusion Sensitivity Profile converts sensitivity analysis results into a structured report that identifies which analytical variations matter most —and communicates that clearly to both the author and the reviewer.

STEP 1 —The Pitfall

Sensitivity analysis output is often a table of numbers. The table tells you that some configurations produce different results from others, but it does not tell you which differences matter, which analytical choices are driving the fragility, or how to communicate fragility to a reader without undermining confidence in the result.

Without a structured summary, sensitivity analysis results are either ignored (if they support the conclusion) or buried (if they reveal fragility). The Conclusion Sensitivity Profile is designed to make sensitivity results actionable and publishable.

[Image placeholder: A structured report card with: Robustness rating (High / Medium / Low / Fragile), a breakdown of which analytical dimension drives the most variation (outlier handling: high impact; test choice: low impact), and a recommendation for disclosure language based on the fragility pattern.]

Conclusion Sensitivity Profile report card: robustness rating, driver dimensions, and recommended disclosure language.

STEP 2 —Journal Requirement

While journals do not explicitly require a "sensitivity profile," they do require that sensitivity analyses be reported with enough detail to evaluate the robustness of primary conclusions. A structured summary satisfies this requirement more efficiently than a raw table of numbers and makes the author's interpretation transparent.

STEP 3 —Licklider Solution

Input

Sensitivity analysis grid from Sensitivity Analysis Engine (4.2.1)
Primary conclusion definition (significance, direction, minimum effect)
Sensitivity dimensions tested and their variations

Output

Robustness rating: High (>85% of configurations consistent), Medium (70—5%), Low (50—0%), Fragile (<50%)
Driver identification: which dimension (outlier handling, test choice, etc.) accounts for the most conclusion variation
Direction sensitivity: does the conclusion change sign in any configuration?
Disclosure language auto-generated based on robustness rating and driver dimensions
Recommendation: if Fragile, suggests replication or pre-registration before high-stakes publication

Guard

If the robustness rating is Fragile (primary conclusion reverses in >50% of configurations), the guard issues a blocking warning for publication-ready export, requiring the user to either: (1) switch to a more robust analytical choice as the primary, or (2) add the Conclusion Sensitivity Profile as a mandatory disclosure in the export package.

[Image placeholder: Full Conclusion Sensitivity Profile report showing: Robustness = Medium (74%), primary driver = outlier handling (causes 4 of 5 sign changes), and auto-generated disclosure text: "The primary conclusion was consistent across 9 of 12 tested configurations. Sensitivity to outlier inclusion/exclusion was the primary source of variation."]

Full Conclusion Sensitivity Profile with robustness rating, driver analysis, and auto-generated disclosure text.

STEP 4 —Draft Output (Draft / Needs review)

Robustness assessment (Conclusion Sensitivity Profile): The primary
conclusion (Group A significantly higher than Group B) was consistent
across 9 of 12 tested analytical configurations (robustness rating: Medium,
74%). The primary driver of variation was outlier handling: configurations
including all data points produced p-values ranging from 0.041 to 0.089,
while configurations with pre-declared outlier exclusion (ROUT Q=1%)
produced p-values of 0.007—.031. All configurations showed a consistent
direction of effect (Group A > Group B).