Fix Bug Of Back Data Not Being Used By Link Filter.

Author: Piwington

statistical_methods_library/imputation/ratio_of_means.py

@@ -219,11 +219,6 @@ def impute(
         ["target", "forward", "backward", "construction"],
     )
 
-    if link_filter:
-        filtered_refs = input_df.filter(link_filter).select(
-            col(reference_col).alias("ref"), col(period_col).alias("period")
-        )
-
     # Cache the prepared back data df since we'll need a few differently
     # filtered versions
     prepared_back_data_df = None
@@ -232,6 +227,12 @@ def impute(
         prepared_back_data_df = validation.validate_dataframe(
             back_data_df, back_expected_columns, type_mapping, ["ref", "period"]
         )
+        input_df = input_df.unionByName(back_data_df, allowMissingColumns=True)
+
+    if link_filter:
+        filtered_refs = input_df.filter(link_filter).select(
+            col(reference_col).alias("ref"), col(period_col).alias("period")
+        )
 
     # Store the value for the period prior to the start of imputation.
     # Stored as a value to avoid a join in output creation.