|MAISY RECS Hourly Loads & Emissions Databases
Question 8. Can I Use MAISY RECS Data for Substate Geographic Analysis?
Short Answer: Yes. Just make sure that geographic segments include enough households
to provide a representative sample. Geographic segments
should be no smaller than about 100-150 household records. Smaller geographic/weather-based
segments may not provide reliable estimates of variables of interest. We also suggest that
at least 20 household are provided for individual variables of interest.
Longer Answer: The Residential Energy Consumption Survey (RECS) sample was designed to provide a statistically representative sample of all households in each state. RECS sample design takes into account the geographic diversity that exists in each state because geographic variables correlate with space heating and air conditioning energy use. We recommend that geographic segments include at least 100-150 household records with at least 20 households for individual variables of interest.
However, it is important to remember that RECS Databases are surveys so any results that are calculated from database variables include some sampling error as is the case with any survey database. Calculating variable averages, say average lighting/TV kWh use, from the RECS database will include a +/- margin of error as is the case with any survey database.
Calculating variable averages from smaller geographic samples can result in unacceptably large sampling errors. One way to judge the extent to which a smaller sample is contributing to larger sampling errors is to calculate the variable average for different sample sizes. For example if a variable average based on a sample of 20 households is significantly different than one calculated with a sample of 30 households, but a sample of 40 households provides an estimate similar to that of the 30 household sample one can infer that the 20 household sample in not representative of the population and the variable average should be based on that provided by the 30 household sample. While MAISY RECS Databases support general drill-downs such as segmenting 7 am household space heating peak kW by dwelling unit floor space, multiple segmentation including 5 floor space categories and 5 income categories may provide segments with too few households in each segment to be reliable estimates.
For example, if the northern/coastal California subsample has 250 household records but the variable of interest is electric space heating which includes only 25 households, average energy use, hourly loads and other variables for these households are not likely to be “statistically reliable.”
As an alternative to applying a MAISY RECS Databases for substate regions with limited sample households, consider standard MAISY Residential Energy Use and Hourly Loads Databases that provide a sample of 6.7 million residential households across the US rather than the 18,400+ households in the MAISY RECS databases. Standard MAISY Residential Energy Use and Hourly Loads Databases also provide ZIP-level geographic detail compared to the RECS state level detail.
Related MAISY Topics