RELEASE NOTES 

		spatstat 1.9-0

	 	3 May 2006

We thank Colin Beale, Matt Reiter and Malte Spiess for contributions.


OVERVIEW

	Version 1.9 of spatstat will include 
		- improvements to the simulation code
		- more methods for statistical inference
		- more support for pixel images and line segment patterns
		
	In the current version 1.9-0 we have fixed several bugs in rmh,
	accelerated some of the simulation algorithms, and added
	some new simulation routines. There is some new code for 
	classical statistical inference, and more support for pixel images.
	
NEW FUNCTIONS

	quadrat.test
	    	Performs a chi-squared goodness-of-fit test 
		of the Poisson process (homogeneous or inhomogeneous), 
		based on quadrat counts.
		
	vcov.ppm
		Variance-covariance matrix for parameter estimates
		in a fitted Poisson point process model.

	rstrat
	rsyst
	rcell
		Functions to generate random point patterns:
		rstrat for stratified random,
		rsyst for systematic random (displaced grid),
		rcell for simulation of the cell process

	nnwhich
		Finds the nearest neighbour of each point in a 
		point pattern.

	nncorr
		Nearest neighbour correlation of a marked point pattern

	cut.im
	quantile.im
	as.matrix.im
		Methods for 'cut', 'quantile' and 'as.matrix'
		for pixel images.

	levelset
		Level set of a pixel image

	solutionset
		Evaluates a logical expression involving pixel images
		and returns the region where the expression is true


IMPROVEMENTS/ENHANCEMENTS

	envelope
		Now also calculates simultaneous critical bands
		(global=TRUE) and allows transformations of the
		summary statistic (transform=expression(.....)),
		for example the L-function.

	as.im
		Now also accepts data in the format list(x,y,z)
		expected by the standard R functions image.default 
		and contour.default.

		Note: this is also useful for converting data 
		returned by the functions 'trmat' and 'prmat'
		in the 'spatial' package.

	quadcount
		The return value has attributes "xbreaks" and "ybreaks"
		specifying the boundaries of the quadrats used in the counting.

	setcov
		Now has arguments to control pixel resolution.

	markcorr
		Calculation is now much faster.
	
	rmh
	rmpoint
		Now run slightly faster.

	ppm
		ppm now checks for, and warns about, NA's in the covariates,
		which may arise when the domain of definition of the covariates
		does not cover the observation window of the point pattern.
		
	lurking
	diagnose.ppm
		now permit NA's in the residuals.

	data(residualspaper)
		Example code in residualspaper$plot()
		updated in accordance with updates to spatstat.

	Documentation
		help files improved.
		demo(spatstat) improved.

	Citation
		Typing citation("spatstat") now prints the appropriate 
		references for citing spatstat in publications.

BUGS

	[1] rmh
		[Spatstat version 1.8-9 only]
		Version 1.8-9 contained a cut-and-paste error
		which caused rmh to produce completely incorrect output
		in all cases except the Poisson process. 
		Spatstat version 1.8-9 was withdrawn from CRAN
		after 2 days on the website. Any simulations performed 
		in version 1.8-9  are probably incorrect. 
		The bug is now fixed.

	[2] rmh
		rmh produced incorrect output when simulating a non-Poisson,
		*multitype* point process with a *nonstationary* trend
		when there were *different trends* for each type of point 
		(equivalently, the trend depends on both location and marks).
		Fixed.

	[3] rmh
		rmh crashed if the initial state was empty
		(i.e. if rmhstart$n.start = 0 or rmhstart$x.start$n = 0).
		Fixed.

	[4] rmh
		When rmh.default reported that it was "increasing the 
		storage space and continuing", its behaviour was very slightly
		incorrect. The most recently-added point in the simulation 
		algorithm was deleted. This is strictly speaking incorrect,
		but it is very unlikely to have had any measurable effect on 
		simulation results.

	[5] rmh
		when control$p = 1 and start$x.start is given, 
		rmh ignored control$fixall. If control$fixall=TRUE,
		rmh was supposed to generate patterns with the same number
		of points of each type as in start$x.start, but this did not
		occur. It has now been fixed.

	[6] "[.fv"
		Did not correctly extract subsets of the columns.
		Fixed.

	[7] markcorr
		The default bandwidth 'bw' was inappropriately small in 
		the case method="density".
		Fixed.

	[8] markcorr
		The default smoothing is inappropriate in 
		the case method="smrep". This cannot easily be fixed.
		Manual control of the smoothing parameter hmult is required.

	[9] markcorr
		The algorithm for the border correction (correction="border")
		was mathematically incorrect and cannot be repaired. 
		This option has been deleted from the code. 

	[10] default.dummy
		some dummy points were occasionally created outside the 
		window of observation, when the window was a mask.
		Fixed.