Many common software questions are answered in the categories below. If you can’t find a response for yours, try checking our User Forum, or, contact HelpDesk for direct assistance

Metrix » Tuning
Q: What do Net Mean Bias and Monthly Mean Error mean?


The two terms, \"Net Mean Bias\" and \"Monthly Mean Error\" are simply ways to quantify how closely Metrix\'s baseline model mimics the bills during the tuning period. Both of these concepts come from ASHRAE GPC14P (the guideline committee on Measurement of Energy and Demand Savings). ASHRAE GPC14P will issue a guideline on the measurement of energy and demand savings for public review by the second quarter of 2000. Note that \"Monthly Mean Error\" is referred to in GPC14P by the mouthful \"Coefficient of Variation of the Root Mean Square Error\", or CVRMSE.

During the tuning period, some bills will be above the value predicted by the baseline model, some will be below. Over the course of a year, these errors will tend to cancel each other. The \"Net Mean Bias\" is the accumulated error over the tuning period, expressed as a percentage per year. The Net Mean Bias should be as close to zero as possible. The latest draft of ASHRAE GPC14P requires this number to be zero, but some committee members and public respondents have pointed out the impracticality of this. The Net Mean Bias is the total error for the entire tuning period.

The Net Mean Bias is calculated using the equation:

Net Mean Bias = (Total Base year Usage - Total Calculated Baseline) / Total Base year Usage

The monthly errors of the baseline model can be large or small, taken individually. If they tend to be small, one would say that \"bills cluster closely around the baseline model.\" If they are large, one might say \"there is random model error\". The \"Monthly Mean Error\" is a quantitative expression of how much this random error is.

The Monthly Mean Error is calculated as follows:

  1. Square the monthly error (Baseline - Reading in the meter tuning contract) for each month of the tuning period
  2. Sum the squares of the monthly error
  3. Divide the sum by the number of bills in the tuning period minus the number of tuning variables (2 if just HDD and days, 3 if CDD, HDD, and # of days, for example)
  4. Take the square root
  5. Divide the square root by the average of the total base year usage

Again, smaller is better, but according to ASHRAE GPC14P, 20% can still be acceptable for energy use, and 30% for demand.

However, note that the savings to be measured by this method must be at least twice the monthly mean error. Suppose, for example, that the monthly mean error for baseline model A is 5%. This means that the expected savings must be at least 10% of the baseline, in order for baseline model A to be permissible.

I can’t find the answer to your question ? Try checking our User Forums.

Otherwise feel free to contact our Tech Support staff at (805) 329-6565, or via email at helpdesk@abraxasenergy.com.

MetrixMarketManagerReportMakerOption C