The MCP discontinuity test determines whether a long-term adjustment's onsite dataset has
a discontinuity compared to a blend of the long-term datasets.
The test consists of the following steps:
Calculate the time series of quarterly means for the onsite dataset and each long-term dataset.
From each quarterly mean time series, calculate the time series of quarterly mean differences by subtracting each quarterly mean from the preceding quarterly mean.
Select the five long-term datasets most strongly correlated with the onsite dataset.
From the selected long-term datasets, calculate a weighted mean time series of quarterly mean differences, weighted by the R2 correlation coefficient between long-term dataset and primary onsite dataset.
Construct an onsite adjusted quarterly mean time series by starting from the first onsite quarterly mean value, and then in each subsequent quarter, adding the long-term weighted mean difference value for that quarter.
Construct a time series of the difference between the measured onsite quarterly means and the adjusted onsite quarterly means.
Apply the Eastering-Peterson test to look for a discontinuity in that measured-minus-adjusted time series.
A summary table indicates the significance of the two-phase curve fit and the change in mean at the
most likely discontinuity point. If both significance values exceed the significance level that you
enter, the overall conclusion will be that a discontinuity does exist in the onsite dataset.
Windographer performs the MCP discontinuity test on the Onsite vs. Reference tab.
