Statistics Glossary → Observation window (study region)
Observation window (study region)concept
In point pattern and density analyses, the boundary of the region in which points could be observed and in which the analysis is valid.
In point pattern analysis, where points could occur matters as much as the points themselves. Intensity (number of points / window area), Ripley's K and L functions and simulations of complete spatial randomness (CSR) are all computed relative to the window. If the window is larger than the real study area, the points can look 'clustered'; if it is smaller, 'regular'. In interpolation methods such as kriging, the window limits the area where predictions are made and prevents extrapolation into regions beyond the observations.
When the window is derived from the points, for example as a bounding rectangle or convex hull, it usually underestimates the real area and slightly overstates intensity. If the true boundary of the study area is known (a park boundary, a sampling region), it should therefore be supplied as a polygon. Points near the window boundary have some of their neighbours outside the window, where they cannot be observed; this edge effect is handled by edge-correction methods. Points falling outside the window are dropped from the analysis, and this should be reported.
An everyday example
The locations of 150 trees in a park have been mapped. If the rectangle around the points is used as the window instead of the park boundary, the car park and the lake, where there are no trees, are included in the area. Even if the trees are randomly distributed within the park, they then look 'clustered'. Supplying the park boundary as a polygon avoids this error.
In MerQur
In point pattern analysis, the 'Observation window' option (Bounding box, Convex hull, Rectangle entered manually, Polygon layer) defines the analysis region; points outside the window are dropped and reported. In Variogram + Kriging, 'Prediction area' (Convex hull (mask), Bounding rectangle) sets the region where predictions are made. In the KDE density map, 'Edge correction' (None, Bounding rectangle, Convex hull) inflates the density by the kernel mass that falls outside the observation window.
Analyses where this concept appears
KDE Density MapVariogram + Kriging (Spatial Interpolation)Point Pattern Analysis (Ripley K/L, G/F/J, CSR)