<result> <column_name>Sleep_Quality_Score</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.72</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 a 1-5 rating level 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> The table is a DIMENSION with no sequence axis, so multiplicative is implausible and ordinal_additive is unlikely since there is nothing to normalize along. The column is a bounded 1-5 integer rating with no naming or unit signals for a ratio. Its mean (2.47) and median (3) are far from 1.0, and the negative log_symmetry_gain argues against ratio behavior. It is a posted level or score where summing across entities is meaningless and averaging is the natural rollup, so intensive_additive fits best. Confidence is moderate because the score is ordinal and discrete, so median or distribution reducers may be preferable to the mean. </reasoning> <table_type_conflict>false</table_type_conflict> </result>