Knowledgebase:
EC Express: Statistical Outliers
Posted by Matthew Heinz on 06 June 2012 11:21 AM

Statistical Outliers

Handling Statistical Outliers

Statistical outliers are handled differently in the EnergyCAP Express Calendarization/Normalization process than they are in the EnergyCAP Enterprise Cost Avoidance module.

In the Cost Avoidance module (currently available only in EnergyCAP Enterprise), statistical outliers (those periodic usage data points which fall more than two standard deviations away from the regression line defining a use/weather correlation for a meter) are handled as exceptions and are excluded from the weather adjustment process. In other words, if June is a “strange” month (for whatever reason), its data is excluded from the regression calculation. This allows a better usage vs. weather correlation using the remaining (non-excluded) bills. 

In EnergyCAP Express, the Calendarization and Normalization processes handle outliers differently. These processes are less sophisticated and less calculation-intensive than Cost Avoidance. Outliers are not excluded or handled differently than other bills. If an outlier causes the regression to fail, it invalidates weather adjustments for that meter, that season, that year. If an outlier exists and the regression remains valid, the weather factors are applied to outlier bills just as they are applied to non-outlier bills.

Identifying Outliers

How do we identify outliers? As illustrated in this sample chart from Buildings & Meters > Calendarized Data > Use vs. Weather, an outlier is a bill point that falls far from where it might be expected to fall, based upon the trend established by other bills.

In this example, the June bill for 2006 (usage data point identified by the arrow) is far above the regression line, whereas the other summer bills fall close to the regression line.

Value of Outliers

Outliers provide potentially valuable information because they help identify an abnormal bill that has usage far different than we might expect. Questions that outliers prompt might include:

  • Is the recorded usage accurate?
  • Was the usage entered correctly?
  • Were we billed correctly?
  • Did something unusual occur in the building during the month that caused higher than expected usage?
  • Are there problems with HVAC or lighting controls?
(0 vote(s))
Helpful
Not helpful

Comments (0)
Post a new comment
 
 
Full Name:
Email:
Comments: