Modifiable temporal unit problem

The sensitivity of statistical results to how time is divided into units, including the duration and boundaries of those units.

Connect

Temporal aggregation: The combination of observations from shorter time intervals into larger time intervals. Changing the aggregation interval can change averages, correlations, and apparent trends.
Time series: A sequence of observations indexed in chronological order. The problem arises when such sequences are grouped into analytical time units.
Epidemiology: The study of the distribution and determinants of health and disease in populations. Disease rates and outbreak patterns can change with the time windows used.
Temporal disaggregation: The estimation of higher-frequency observations from data reported at lower frequencies. It addresses coarse temporal data by estimating finer units, rather than studying sensitivity to unit choice.
Temporal resolution: The fineness of the time intervals at which observations are recorded or analyzed. The chosen resolution determines which short-lived patterns remain visible.
Sampling interval: The time elapsed between successive observations in a sampled process. It sets a basic temporal unit before any further aggregation.
Epidemic curve: A graph showing the number of new cases of an outbreak over time. Its shape and apparent peaks depend on the temporal bins used for case counts.
Change-point detection: Methods for identifying times when the statistical properties of a sequence change. Detected change points can shift when observations are grouped into different intervals.
Time-series aggregation: The transformation of time-series observations into summaries over specified intervals. It makes explicit how interval length affects statistics calculated from the same series.
Statistical aggregation: The summarization of multiple observations into a smaller set of statistical values. Temporal units determine which observations are combined in each summary.