RainbowStats command reference

ACF_CHART

Displays the autocorrelation of a time series across successive lags, helping reveal persistence, seasonality, and remaining structure in model residuals.

Syntax

ACF_CHART(series)
series

The numeric time series whose autocorrelation function will be plotted.

What the chart shows

For each lag, the chart measures the correlation between the series and a delayed copy of itself. Autocorrelation values range from −1 to +1. Bars near zero indicate little linear dependence at that lag; larger positive or negative bars indicate stronger serial structure.

Slow decay

Often signals persistence or a non-stationary level series.

Repeating peaks

Can reveal a seasonal or cyclical pattern.

Residual spikes

Suggest that a fitted model may still be missing time dependence.

Examples

Recommended for unemployment-rate changes ACF_CHART(logdiff(UNRATE))
Inspect the level series ACF_CHART(UNRATE)
Inspect stock-market returns ACF_CHART(logdiff(DSG10))

Interpretation guidance

Use stationary data when possible.

Trending economic levels often produce high autocorrelation even when the apparent relationship is not economically meaningful. Apply a transformation such as diff(series) or logdiff(series) when appropriate.

  1. Plot the original series and look for trend or changing variance.
  2. Transform the data if necessary, then run ACF_CHART.
  3. After fitting a model, run the chart on its residuals.
  4. Well-specified residuals should generally resemble uncorrelated noise.

Notes

Autocorrelation is descriptive, not proof of causation. Individual spikes should be interpreted alongside the sample size, the number of lags examined, the data frequency, and the economic context.