One Row of Data is Not One Real House

When working with raw ResStock data, it may be easy to assume that each line of data or each file represents one real U.S. house. However, this is not the case. Instead, each row of data in a ResStock metadata file has one sampled dwelling unit that represents a segment of the housing stock with the same characteristics. A row of data is just representative and cannot be tracked or linked to an actual building or address.

For example, in the illustrative table below with a subset of ResStock metadata, the first few columns are labeled bldg_id and weight.

bldg_id	weight	in.city	…	in.geometry_building_type	in.geometry_floor_area
186	242.13101	AL, Huntsville	…	Single-Family Detached	2500-2999
212	242.13101	AL, Mobile	…	3 or 4 unit	4000 +
239	242.13101	AL, Birmingham	…	Mobile Home	750-999

bldg_id is the identification number of the dwelling unit simulated in ResStock. weight is the sample weight, or the number of real dwelling units in the U.S. housing stock that the individual bldg_id represents. In the table above, bldg_id 186 has a weight of 242.13101, meaning that this simulation represents approximately 242 real single-family detached homes in Alabama. With ResStock’s typical quota-based sampling approach, each ResStock run equally weights all housing units. For this example, each of the 550,000 sampled housing units in ResStock represents about 242 real housing units each. The weight field is included in all ResStock data files.

The other column titles such as in.city, in.geometry_building_type, and in.geometry_floor_area highlight different characteristics about the line of data. in.city represents State, city so the first line represents Alabama, Huntsville. in_geometry_building_type is the geometry building type and the format in. represents that it is in an input. in.geometry_floor_area tells the square footage of the floor area of the dwelling unit. Descriptions of each column name can be found in the data_dictionary files in all ResStock datasets.

For a comprehensive solution that limits uncertainty to about 15%, we recommend using at least 1,000 samples (representing about 242,000 real dwelling units) in this example. More guidance on this topic, can be found in this explanation.

An additional point to emphasize is that Resstock models dwelling units, not entire buildings. For example, in the table above, bldg_id 212 says the in.geometry_building_type is 3 or 4 unit. That means the row of data represents one dwelling unit or one apartment within a building that has 3 or 4 units total. The point is espectially important for analyses that focus on multifamily buildings or for regions that have a high proportion of multifamily buildings.