When adjusting schemas after prediction failures…

When adjusting schemas after prediction failures, distinguish between calibration adjustments (schema correct in structure, wrong in parameters) and structural revisions (missing variables or wrong causal model).