RELEASE NOTES 

		spatstat 1.15-3

		14 May 2009

We thank Julian Burgos, Michael Drinkwater, Ege Rubak,
Eva Stehle and Marcelino de la Cruz for contributions.

OVERVIEW
	New summary functions for marked point patterns.
	Connected component transform for images and binary masks.
	Improvements to `distmap' for polygons.
	Extra functionality for density.ppp and smooth.ppp.
	New dataset from astronomy. Bug fixes.

NEW DATASET

	shapley
		Point pattern of galaxies in an astronomical survey

NEW FUNCTIONS

	connected
		Connected component transform of a pixel image
		or binary mask.

	markmean
	markvar
		Spatially smoothed estimates of the mean mark
		and the variance of marks, in a point pattern
		with numeric marks.

	nnmean
		Nearest-neighbour mean mark index

	nnvario
		Nearest-neighbour mark correlation index


IMPROVEMENTS/ENHANCEMENTS

	distmap.owin
		distmap.owin now produces a much more accurate result
		for polygonal windows. 
		It also has two new arguments 'discretise' and 'invert'. 

	incircle
	dilate.owin 
	erode.owin 
		These commands now handle polygonal windows more accurately.

	density.ppp
		New argument 'at' allows computation of values of
		kernel smoothed intensity at the data points,
		without pixel discretisation.

	smooth.ppp
		New argument 'at' allows computation of exact values of
		interpolated function at the data points.

		
	Kdot, Kcross, Ldot, Lcross
		The option 'correction="none"' is now implemented.


	ripras.R
	ripras.Rd
		ripras has new argument 'f' controlling expansion of window

BUGS

	[1] quadrat.test.splitppp 
		This function did not pass the "..."
		arguments to quadrat.test.ppp. 
		Fixed.

	[2] Kest
		If the argument 'domain' was given, then
		'correction="none"' did not work.
		(Detected by Eva Stehle)
		Fixed.

	[3] Kest
		If the argument 'domain' was given,
		the return value was incorrectly labelled.
		Fixed.

	[4] plot.ppp
		Crashed with an error about 'unimplemented pch value'
		if the point pattern was multitype with more than 25 types.
		(Spotted by Marcelino de la Cruz)
		Fixed.

	[5] kstest
		for some datasets, the p-value had excessive random variability
		because the procedure used for jittering the covariate
		was ill-chosen. (Spotted by Ege Rubak)
		Fixed.