After the regression analysis is completed, a CUSUM Analysis can be carried out on that same Dependent Variable.  The steps the user must take to carry out the analysis are as follows:

1. Select the same Dependent Variable used from the regression completed, from the drop-down list, or use the data from other regressions carried out which have already been tabulated.
2. Enter the values for the Intercept and for the Independent Variables used in the regression carried out.
3. The values for the Dependent Variable selected appear in the table provided on the worksheet.
4. All of the data for the Independent Variables chosen are calculated using the product of the coefficient values and equivalent variable values for that time period.

For example, January 2004:

Independent Variable (1): Degree Days

= Coefficient (1)*(Degree Day value in January 2004)

1. The predicted target consumption will be calculated automatically in the table.
2. The Difference between the actual consumption of the Dependent Variable and the Predicted Consumption is calculated automatically.
3. The differences calculated are then summed cumulatively and tabulated automatically.
4. A CUSUM graph will then be plotted at the bottom of the worksheet.

The CUSUM graph plotted indicates the performance variation of the building over the time period being analysed.  The gradient of the slope highlights the level of efficiency of the building.  A slope going up over time shows a poor building performance; a slope going down indicates good performance; and if a more horizontal slope develops over, time this indicates that there is a change in performance; which can be both good or bad.

The points on the CUSUM graph where a change in performance occurs can be explained by referring back to the installation dates of small scale energy saving methods; which have been entered by the user in the Main Page worksheet.

Consumption Predictions     

If a linear slope is plotted on the CUSUM graph at the end of the period being analysed, a further regression can be carried out for the same variables, over the relevant period, using the same steps described prior. 

The results from this regression analysis i.e. Intercept, Coefficient values, can be used to produce a new performance equation.  This equation can be used to predict the consumption for the forthcoming months and compared to the actual consumption being metered.  If a large difference is noticed between these two values, this may indicate that the performance of the building is being affected by some unknown variable e.g. the standard of staff good practice is falling; a control system in the building is faulty.