```xml <result> <column_name>Previous</column_name> <family>additive</family> <subtype>intensive_additive</subtype> <confidence>0.65</confidence> <requires_user_confirmation>true</requires_user_confirmation> <sum_meaningful>false</sum_meaningful> <signed>false</signed> <reducers> <primary_reducer>mean</primary_reducer> <valid_reducers>mean, min, max, median, mode</valid_reducers> <excluded_reducers>sum (intensive level, summing 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> Table-type gate: STATE with no sequence axis makes multiplicative highly implausible. The column name "Previous" suggests a prior state or level value, not a ratio or rate. The dtype is object (likely categorical or ordinal codes), with very low distinct_fraction (0.65%) indicating discrete levels rather than continuous measurements. The wide range (0.0 to 1791.0) with sample values at round powers of 10 suggests ordinal category codes or discrete state identifiers. Treating as intensive_additive (a categorical level) where mean represents the average category code, though this may actually be a categorical dimension requiring mode rather than numeric aggregation. Confidence reduced due to ambiguity between ordinal codes and true intensive measure. </reasoning> <table_type_conflict>false</table_type_conflict> </result> ```