10  K-function

10.1 Motivation

The PCF \(g(r)\) describes how interaction between points varies with distance.

However, \(g(r)\) is a local measure - it focuses on behavior at specific distance \(r\).

In many situations, it is useful to have a summary that captures the total interaction up to a given distance.

This leads to the \(K\)-function.

10.2 Definition

The \(K\)-function, denoted \(K(r)\), is defined such that

\[ \lambda K(r) \]

is the expected number of further points within distance \(r\) of a typical point.

The \(K\)-function answers the question:

Given a point of the process, how many other points do we expect to find within distance \(r\)?

More precisely:

  • \(K(r)\) describes the cumulative interaction up to distance \(r\).
  • It aggregates behavior over all distances from \(0\) to \(r\).
Note

Just remember:

  • \(g(r)\) tells us what happens AT distance \(r\)
  • \(K(r)\) tells us what happens WITHIN distance \(r\)

10.3 Relation to the Poisson process

For a Poisson process with intensity \(\lambda\), we have:

\[ K(r) = \pi r^2 \]

Note

No explanation why??????? Are you fucking serious dog

This corresponds to the area of a disc of radius \(r\), reflecting complete spatial randomness.

Note

Shitty ahh explanation.

10.4 Detecting interaction

Comparing \(K(r)\) to \(\pi r^2\) reveals the spatial structure:

  • \(K(r) > \pi r^2 \Longrightarrow\) clustering

  • \(K(r) < \pi r^2 \Longrightarrow\) repulsion

  • \(K(r) = \pi r^2 \Longrightarrow\) no interaction

10.5 Relationship to the PCF

The \(K\)-function and the PCF are closely related.

In particular, the \(K\)-function can be expressed as

\[ K(r) = \int_0^r 2\pi s\ g(s)\ ds \]

Thus:

  • the PCF \(g(r)\) describes local interaction
  • the \(K\)-function aggregates this into a cumulative measure.
Note

The \(K\)-function is essentially a smoothed, accumulated version of the PCF.

10.6 Advantages and Limitations

Advantages

  • Easier to estimate in practice (less noisy than the PCF)

  • Provides a stable summary of interaction (is this not just the same as the last point).

  • Widely used in exploratory analysis

Limitations

  • Being cumulative, it can mask local features

  • Different processes can produce very similar \(K(r)\)

  • Interpretation is less direct than for \(g(r)\)

10.7 Summary

  • The \(K\)-function describes the expected number of neighbours within distance \(r\)
  • It provides a cumulative measure of interaction
  • It is related to the PCF via: \[ K(r) = \int_0^r 2\pi s \, g(s)\, ds \]
  • Comparing \(K(r)\) to \(\pi r^2\) reveals clustering or repulsion

10.8 Next Step

We now consider how second-order quantities such as the PCF and \(K\)-function can be estimated from observed data.