enriching the reported results

[azzashoaibi]

The current outputted results can be improved by addressing the following issues:

1. Statistical Test Consistency in Part 2 Results

• Currently, both paired t-tests and Wilcoxon tests are used based on normality checks.
  - Proposal: Use Wilcoxon test consistently across all arms for better methodological consistency.

2. Non-Inferiority Threshold Clarification

   • The non-inferiority p-value and statistical test rely on a threshold that is currently unspecified.
   • Action Needed: Define and document this threshold—suggested value is 10%

3. Missing Weighted F-Score Table in Part 2 Results

• The report lacks weighted F-score results per phenotype and per AI arm (the results are across all phenotypes)
• Action Needed: Add a table showing weighted F-scores for each phenotype and AI arm.

4. Include Unweighted F-Score Results
   - Unweighted F-score results are not currently included.
   - Action Needed: Add these to the report for completeness.

5. Agreement Metrics in Part 1 Results**
   Agreement is currently pooled across all AI arms.
   Action Needed::
   - Report agreement (overlap bars) separately for each AI arm and each disease.
   - Weight these bars by concept prevalence.

6. Precision and Recall Reporting**

   • Only F-score is reported.
   • Action Needed : Report precision and recall separately for each AI arm and each disease.