def modify_dataset(ds: MultiRegionDataset) -> MultiRegionDataset:
    ts_copy = ds.timeseries.copy()
    # Test positivity should be a ratio
    ts_copy.loc[:, CommonFields.TEST_POSITIVITY_7D] = (
        ts_copy.loc[:, CommonFields.TEST_POSITIVITY_7D] / 100.0)

    levels = set(
        Region.from_location_id(l).level for l in
        ds.timeseries.index.get_level_values(CommonFields.LOCATION_ID))
    # Should only be picking up county all_df for now.  May need additional logic if states
    # are included as well
    assert levels == {AggregationLevel.COUNTY}

    # Duplicating DC County results as state results because of a downstream
    # use of how dc state data is used to override DC county data.
    dc_results = ts_copy.xs(DC_COUNTY_LOCATION_ID,
                            axis=0,
                            level=CommonFields.LOCATION_ID,
                            drop_level=False)
    dc_results = dc_results.rename(
        index={DC_COUNTY_LOCATION_ID: DC_STATE_LOCATION_ID},
        level=CommonFields.LOCATION_ID)

    ts_copy = ts_copy.append(dc_results, verify_integrity=True).sort_index()

    return dataclasses.replace(ds,
                               timeseries=remove_trailing_zeros(ts_copy),
                               timeseries_bucketed=None)
def add_state_location_id_index_level(df: pd.DataFrame) -> pd.DataFrame:
    """Returns df with the location_id of the state in a new level STATE_LOCATION_ID."""
    state_index = (df.index.get_level_values(
        CommonFields.LOCATION_ID).map(lambda loc_id: Region.from_location_id(
            loc_id).get_state_region().location_id).rename(STATE_LOCATION_ID))
    return df.set_index(state_index, append=True)