<result> <column_name>PetalWidthCm</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.68</confidence> <requires_user_confirmation>false</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, min, max, median, distribution</valid_reducers> <excluded_reducers>sum (summing per-entity petal widths across flowers has no physical meaning), product (not a ratio), geomean (not a ratio)</excluded_reducers> <average_semantics>arithmetic_mean</average_semantics> </reducers> <inner_outer_measure> <inner_measure_reducer>min</inner_measure_reducer> <outer_measure_reducer>max</outer_measure_reducer> </inner_outer_measure> <multiplicative_partner_derivable>false</multiplicative_partner_derivable> <reasoning> The table is a DIMENSION with no sequence axis, so multiplicative is very unlikely and no reconstruction signals are available. The name "PetalWidthCm" has no ratio tokens, and the "Cm" suffix indicates an absolute physical length. Center-of-mass does not support a ratio: the median is 1.3, min is 0.1, max is 2.5, center_near_one is false, and log_symmetry_gain is negative. The column is a per-entity physical attribute (one row per flower), so the natural rollup is a mean rather than a sum. I chose intensive_additive over real_additive for that reason. Confidence is moderate because length-like names can also be treated as extensive. </reasoning> <table_type_conflict>false</table_type_conflict> </result>