Data is represented over a continuous time interval, between the first and last date-time values given on object creation. Each data value therefore represents the value over a specific time interval within this range. These time-intervals are contiguous in the time-domain i.e. the ending point for one interval will be the starting point for the next, and each interval is of equal size. Conceptually, you can think of the x-axis of a graph being segmented into equal portions, with the number of portions being the number of data values. Each value represents the data for its portion of time, which corresponds to a single point in the data set. Such data values can then be used as discrete points on a graph. For example, each pixel might represent one time-interval i.e. one point of data.
Since data values are gathered at specific times, a data summary would be unnecessary if each time-interval contained exactly one data value. However, this is frequently not the case. Sometimes there will be more than one data value in a time interval, and sometimes there will be no time values in an interval. In the latter case, the value Double.NaN is used as the internal representation for the lack of a value in the time interval corresponding to a point.
When more than one value exists in a time interval, the values for the interval must be grouped together to give a single value in the summary, a practice called "aggregation". Two aggregation policies are presently defined:
When there are no data values in a time interval, the values can be inferred from the existing values in surrounding intervals. This practice is called "interpolation". Three interpolation policies are presently defined:
| |