TS Filler Tool

Per the most common business usage scenarios, time series periods are handled as follows:

• “Time-level”: For intervals
• “Date-level”: For intervals >= a day, the period is “period ending.”

The only modification that the macro makes to the incoming data is: a “rounding” is applied to the [yourDateTimeColumn]. It is based on three key elements:

• Interval
• Earliest datetime value in data stream
• Increment

First, all fractions of the chosen interval are truncated off. For example, if you choose “Hour” as the interval, the macro will truncate any minutes and seconds from the datetime value, so e.g. 1:58pm will become 1:00pm.

After truncating the fractions of the chosen interval, the macro then determines what the valid periods are. The earliest datetime value entering the macro forms the basis of the series. Continuing the above “hour-series” example, if the earliest datetime in the [yourDateTimeColumn] was 1:58pm, then the first period is 1:00pm. If the chosen increment is 5, the subsequent periods are generated based on successively adding the chosen increment—5 hours—to the previous period, resulting in this case to 1:00pm, 6:00pm, 11:00pm etc.

Finally, now that all valid periods have been determined, all datetime values are then adjusted to the appropriate valid period. So a raw value of 5:59pm in the [yourDateTimeColumn] would become 1:00pm, 6:02pm would become 6:00pm, etc.

After all datetime adjustments have been applied by the macro, the resulting [yourDateTimeColumn] is now effectively a “label” for that period. However, the datatype is a date or datetime column (rather than a string), and as such this “label” may not be entirely intuitive. The key is to recognize that any portion of the date or datetime “label” that is “fractional” to the chosen interval should be ignored. So if you interval is “year”, than an example period could be “2018-01-01” but it’s the year that matters in the label; the fractional portions of the label—the “01-01”—should be ignored, meaning that the period is simply the year 2018, or “any time in 2018”. See table below.

Note that, aside from the datetime “rounding” described above, the macro does not alter any incoming data. It doesn’t do any combining, summarizing, uniquing or any type of validations on the incoming data. However, because of the date rounding, multiple measurements within the same period will now carry the same period label. As such, the data is now cleaner and can more readily be summarized because data can be “grouped by” this consistent, rounded period label.

Note that the macro does not account for Daylight Saving Time, so if you have time-level series data that crosses a Daylight Saving Time cutover, it’s possible that the macro will generate an hour row (or minute rows) representing an invalid hour. For example, areas that “Spring forward” one hour in March will have row(s) generated for the 2am hour on the Spring forward day even though that hour never actually occurred. (The clock switches from 1:59:59 to 3:00:00.) This hour can easily be removed with a filter tool after the macro.

To clarify, the following table lists the interval options available, followed by what datatype is returned, and whether the period for that interval type is a “period starting” or “period ending” interval. The table then displays an example increment and “first raw value” in the series, to illustrate what the resulting period “Label” would be and an English description of what the period label actually means. Finally, the table notes which portion of the period label should essentially be ignored, i.e. the portions that are fractional to the interval.