|Metrix » Tuning|
An important assumption behind ordinary least square regressions is that all independent variables are normally distributed. Unfortunately, heating and cooling degree-days have an over-representation of zero or near-zero values during mild and off seasons (e.g., no heating degree-days in summer). In other words, a typical year will include a significant number of electricity bills whose weather-related portion is zero or very small.
To minimize the resulting statistical error it is convenient to exclude from the regression all utility bills corresponding to periods with degree-days below a Degree-Day Threshold. The Degree-Day Threshold can be expressed as average degree-days per day, corresponding to a temperature difference. A typical value is around 2 F (1.1 C) which signifies that any 30-day utility bill where the corresponding degree-days are below 60F-days (33.33 C-days) is excluded from the regression. .... The value of the Degree-Day Threshold can be varied such as to exclude only bills with small, yet most uncertain cooling components.
Such exclusion is usually necessary only if cooling or heating degree-days are the only variable in the regression. For multi-variable regressions where at least one variable is not weather-related no bills need to be excluded.
Otherwise feel free to contact our Tech Support staff at (805) 329-6565, or via email at email@example.com.