df.rolling(): Manual

The df.rolling() function is a function (method really) of the DataFrame object which allows us to compute statistics over a moving window of fixed size and fixed step.

A rolling window call returns a Rolling object, not a Series or DataFrame. It becomes a Series or DataFrame only after we apply a function such as mean(), sum(), or apply().

The main parameters are:

• window: the size of the rolling window

• min_periods: minimum number of non-NA observations required to produce a value/statistic

• closed: which side of the interval is closed (valid strings: "left", "right", "both", "neither")

• step/size: the step size with which we roll the window

# Example: rolling mean of a temperature column
rolling_mean = (
    df["temperature"]
    .rolling(window=24, min_periods=12, step=2)  
    .mean().dropna()                            
)