stzGeoProcess

(no comment in the source: the name reads as a sentence) initno extension

what it ISno extension

(no comment in the source: the name reads as a sentence) contentno extension

DOES THIS PROCESS PLACE POINTS WITHOUT REGARD TO ONE ANOTHER? The three that do are the honest null models -- whatever a statistic sees in their output is chance. The four that do not have an interaction built in, and are what a caller tests FOR.no extension

(no comment in the source: the name reads as a sentence) is clusteringno extension

(no comment in the source: the name reads as a sentence) is inhibitingno extension

the parameters, under the names they have in the subject ------------- points per square kilometre -- :Poisson, :MaternIIQCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) intensityno extension

how many points exactly -- :Binomial, :SSIQCSXTZZZIBWU

clusters per square kilometre -- :MaternCluster, :Thomas. The parents are NEVER DRAWN: they are where the clusters are, not points of the pattern, and a reader who saw them would be looking at a mechanism rather than at data.QCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) set mean childrenQCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) mean childrenno extension

the radius of a Matern cluster's disc, kmQCSXTZZZIBWU

the scale of a Thomas cluster's Gaussian, km. It is NOT a radius: about a third of the children land beyond it and a few land far out, which is the whole difference from Matern and is a claim about the mechanism.QCSXTZZZIBWU

the distance no two points may come within -- :MaternII, :SSIQCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) hard core kmno extension

THE INTENSITY SURFACE of an inhomogeneous process, as a GE7b field. That is not a workaround for the engine taking no callbacks: a kernel density estimated from one pattern IS such a field, so the natural workflow -- read a population surface, then ask whether the cases are clustered beyond it -- is two calls with nothing in between.QCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) set intensity gridQCSXTZZZIBWU

(no comment in the source: the name reads as a sentence) has intensity fieldno extension

For the two CLUSTER processes this is the count BEFORE the window clips it. Children thrown outside are lost, and how many depends on how far the mechanism scatters them -- measured on a one-degree square, a Matern cluster of radius 8 km loses 6 per cent and a Thomas of sigma 4 km loses 17, because a Gaussian has no edge. The loss is real and it is the window's, not the process's.no extension

ONE PATTERN FROM THIS PROCESS, inside a window, as a flat lon/lat list.QCSXTZZZIBWU

...and the same as a stzGeoPoints, ready to be measuredno extension

THIRTY-NINE IS THE CLASSIC COUNT, and it is not a floor to be raised for free precision: the band WIDENS with the number of simulations, so an envelope of 999 is a stricter test than an envelope of 39 and the two are not comparable. Being the most extreme of forty is a one-in-forty event under the null -- 2.5 per cent on each side. The count belongs in whatever gets reported.QCSXTZZZIBWU

WHERE AN OBSERVED PATTERN ESCAPES THIS PROCESS'S BAND, and which way: -1 below, 0 inside, +1 above. A reader wants the RANGE OF SCALES at which a pattern is unusual, not one verdict -- a wood can be clustered at ten metres and regular at fifty, and a single p-value cannot say that.no extension

THE VERDICT IN WORDS, with the scales it holds at -- which is the only honest summary, because "clustered" without a range is not a finding.no extension