Skip to content

Home / Statistical Tools / Analysis Tools / Time Series / Partial Autocorrelation

Partial Autocorrelation

Measure the correlation between a series and its own past at each lag, with the influence of the intermediate lags removed.

Where to find it

QXL Stat Tools → Analysis Tools → Time Series → Partial Autocorrelation

Worked example

The Partial Autocorrelation walkthrough builds this analysis step by step from sample data you can copy straight into Excel, and states every number it produces.

All six Time Series walkthroughs are listed on the How-To page.

What it does

The partial autocorrelation function (PACF) answers "how much does lag \(k\) correlate with the present, beyond what lags \(1\) through \(k-1\) already explain?" Where the ACF shows total dependence, the PACF isolates each lag's direct contribution, which is the standard way to choose the order of an autoregressive model.

Options

The dialog is Time Series Correlation, with a Data tab and a Correlation Options tab.

Data tab

  • Excel / GroupBy: analyze flat Excel columns, or split by one or more GroupBy columns
  • Data Columns: one or more numeric columns; each produces its own PACF

Correlation Options tab

  • Auto / Manual lag type: Auto lets Quantum XL pick how many lags to compute; Manual uses your lag count
  • Lag Value: the number of lags when Manual is selected. Default 1; allowed 1 up to the number of observations minus 1.

Combining into a single chart is not available for the correlation tools.

Output

A worksheet named PACF ... containing:

  • A needle-style chart of the PACF against Lag, with confidence bands
  • A data table with Lag, PACF, T-Stat columns

See Also