TimeFrame.rolling_aggregate

TimeFrame.rolling_aggregate(window_size, aggregation_function, columns=None, missing_criteria=None, alignment=RollingAlignment.TRAILING, **kwargs)

Apply a rolling aggregation function to this TimeFrame and return a new TimeFrame with the same temporal structure.

Unlike aggregate(), which reduces the resolution of the data, rolling aggregation preserves the original timestamps: each row in the output corresponds to a row in the input, with the aggregation computed over a sliding window of the specified size.

Parameters:

• window_size (Period | str) – The size of the rolling window.

• aggregation_function (Union[str, Type[AggregationFunction], AggregationFunction]) – The aggregation function to apply.

• columns (str | list[str] | None) – The column(s) containing the data to be aggregated. If omitted, will use all data columns.

• missing_criteria (tuple[str, float | int] | None) – How the aggregation handles missing data. When the actual number of values in a window is below the threshold the result is flagged as invalid via a valid_<column> column.

• alignment (RollingAlignment | str) –

  Where the window is positioned relative to each timestamp:

  • TRAILING (default): window looks backward - (t - window_size, t]. Edge effects appear at the start of the series.

  • LEADING: window looks forward - [t, t + window_size). Edge effects appear at the end of the series.

  • CENTER: window is centered - [t - window_size/2, t + window_size/2]. Edge effects appear at both ends. Not supported for calendar-based window sizes.

  Accepts a RollingAlignment instance or a string ('trailing', 'leading', 'center').

• **kwargs – Parameters specific to the aggregation function.

Return type:

Self

Returns:

A TimeFrame with the same resolution, periodicity, and time anchor as this TimeFrame, containing the rolling aggregation results.