Issue 1140 loss rate rounding (#1170)

* updated loss rate rounding

* fixing a typo in variable name

* fixing typo in variable name

* oops, now ready to push

* updated pickle with float for loss rate columns

* updated a typo, now multiplies all loss rates by 100 consistent with other pcts

* updated with final pickles, all tests passing

* updated incorporating lucas pr comments

* changed literal to field name

data/data-pipeline/data_pipeline/etl/score/constants.py

@@ -69,12 +69,20 @@ PERCENT_PREFIXES_SUFFIXES = [
     "percent",
     "Percentage",
     "Energy burden",
-    "loss rate",
     "greater than or equal to 18 years",
     field_names.PERCENTILE_FIELD_SUFFIX,
 ]
-
 TILES_ROUND_NUM_DECIMALS = 2
+
+# FEMA rounding columns
+FEMA_ROUND_NUM_COLUMNS = [
+    field_names.EXPECTED_BUILDING_LOSS_RATE_FIELD,
+    field_names.EXPECTED_AGRICULTURE_LOSS_RATE_FIELD,
+    field_names.EXPECTED_POPULATION_LOSS_RATE_FIELD,
+]
+
+TILES_FEMA_ROUND_NUM_DECIMALS = 4
+
 # Tiles data: full field name, tile index name
 TILES_SCORE_COLUMNS = {
     field_names.GEOID_TRACT_FIELD: "GTF",

data/data-pipeline/data_pipeline/etl/score/etl_score_post.py

@@ -259,6 +259,16 @@ class PostScoreETL(ExtractTransformLoad):
                     pd.to_numeric(df_100, errors="coerce")
                 ).astype("Int64")
                 df[column] = df_int
+
+            elif column in constants.FEMA_ROUND_NUM_COLUMNS:
+                # Convert loss rates by multiplying by 100 (they are percents)
+                # and then rounding appropriately.
+                df_100 = df[column] * 100
+                df[column] = floor_series(
+                    series=df_100.astype(float64),
+                    number_of_decimals=constants.TILES_FEMA_ROUND_NUM_DECIMALS,
+                )
+
             else:
                 # Round all other floats.
                 df[column] = floor_series(