Optimal Time Offset Window |
The Optimal Time Offset window shifts datasets in time to improve their correlation with each other. To access this window, choose
from the menu.In the drop-down box at the top of the window you specify the 'anchor dataset' to which all the others are compared. The window refers to the other datasets as 'candidate datasets'. The anchor dataset will not move in time, but any of the candidate datasets might move in time.
Windographer evaluates the correlation between the anchor dataset and every candidate dataset, using the Pearson correlation coefficient to characterize the correlation between two datasets. It calculates that correlation coefficient from the primary speed, direction, and temperature data columns, and it does so for many different time offsets to determine which time offset maximizes the degree of correlation.
For each of the candidate datasets, Windographer first resamples the anchor dataset if necessary to match the time context of the candidate dataset, then it shifts the candidate dataset forwards and backwards in time by one, then two, then three time steps and so on, each time calculating the resulting strength of correlation between the two datasets as measured by the Pearson correlation coefficients in the primary speed, direction, and temperature data columns.
A graph of correlation coefficient versus time offset for each candidate dataset appears at the bottom of the window. By default it shows correlation coefficients calculated from speed data because they tend to pinpoint time offsets most reliably, but it can show correlation coefficients calculated from direction and temperature too.
The results in the screenshot below show that in the search for the optimal time offset, each candidate dataset gets shifted forwards and backwards in time by an integer number of that dataset's time steps. Candidate datasets with 10-minute time steps, for example, get shifted forward and backwards in time by 10 minutes, 20 minutes, 30 minutes, and so on, while candidate datasets with 6-hr time steps get shifted by 6 hours, 12 hours, 18 hours, and so on.
The accompanying table shows, for each candidate dataset in comparison to the anchor dataset, and for each for the three data types:
The table also shows the recommended amount of time shift, as the next section describes.
Note that the search for the optimal time shift may proceed farther in one direction than another, and by differing amounts for the different datasets, depending on the time step of each candidate dataset and the patterns that emerge.
Windographer analyzes the results and recommends a time shift for each candidate dataset. This recommendation takes into account the correlation patterns of speed, direction, and temperature, giving top priority to the speed. You can accept or override the recommendations, and when you close the window with OK, Windographer will apply the time offsets you have specified.
Tip: You can offset a single dataset in time with the Apply Time Shift window.
See also
Pearson correlation coefficient