```xml <result> <column_name>referential_integrity</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.85</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 integrity scores is meaningless), product (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> Gate prior: DIMENSION table with no sequence axis makes multiplicative highly implausible. The column name "referential_integrity" indicates a quality score or percentage metric bounded [0,1], which is an intensive level measure, not a ratio that compounds. While center_near_one is true (mean=0.97, median=1.0), this represents high data quality scores clustered near perfect integrity, not a multiplicative factor. The presence of zeros and the bounded [0,1] range confirm this is a normalized quality metric where averaging across entities is meaningful but summing is not. Negative log_symmetry_gain rules out ratio_multiplicative interpretation. </reasoning> <table_type_conflict>false</table_type_conflict> </result> ```