1:
2: \documentclass[a4paper,twoside]{article}
3: \usepackage{amsmath,amssymb,amsfonts}
4: \pdfoutput=1
5: %\usepackage{hyperref}
6: \usepackage{graphics}
7: \DeclareGraphicsExtensions{.pdf, .jpg}
8: \pagenumbering{arabic} \oddsidemargin 0cm \evensidemargin 0cm
9: \textwidth 15cm \topmargin -1cm \pagestyle{myheadings}
10: \markboth{{\small\it Quantifying departures from randomness of the inter-galactic
11: void probablity function}} {{\small\it C.T.J. Dodson} }
12: \parindent 0cm
13: \textheight 24cm
14: \textwidth 17cm
15: \parskip 1mm
16: \newtheorem{theorem}{Theorem}[section]
17: \newtheorem{definition}{Definition}[section]
18: \newtheorem{remark}{Remark}[section]
19: \newtheorem{corollary}{Corollary}[section]
20: \newtheorem{proposition}{Proposition}[section]
21: \newtheorem{example}{Example}[section]
22:
23: \def\R{{\mathbb R}}
24: \def\C{{\mathbb C}}
25: \def\bS{{\mathbb S}}
26: \date{ }
27: \title{Quantifying galactic clustering and departures from randomness of the inter-galactic
28: void probability function using information geometry\footnote{Presented at: Workshop on Statistics of Cosmological Data Sets NATO-ASI Isaac Newton Institute 8-13 August 1999. }}
29:
30: \author{C.T.J. Dodson\\{\small\it School of Mathematics, University of Manchester,
31: Manchester M13 9PL, UK}\\
32: {\small\it ctdodson@manchester.ac.uk} }
33: \begin{document}
34: \maketitle
35: \begin{abstract}
36: A number of recent studies have estimated the inter-galactic void
37: probability function and investigated its departure from various
38: random models. We study a family of parametric statistical models
39: based on gamma distributions, which do give realistic descriptions
40: for other stochastic porous media. Gamma distributions contain as
41: a special case the exponential distributions, which correspond to
42: the `random' void size probability arising from Poisson processes.
43: The random case corresponds to the information-theoretic maximum
44: entropy or maximum uncertainty model. Lower entropy models
45: correspond on the one hand to more `clustered' structures or `more
46: dispersed' structures than expected at random. The space of
47: parameters is a surface with a natural Riemannian structure, the
48: Fisher information metric. This surface contains the Poisson
49: processes as an isometric embedding and provides the geometric
50: setting for quantifying departures from randomness and perhaps on
51: which may be written evolutionary dynamics for the void size
52: distribution. Estimates are obtained for the two parameters of the
53: void diameter distribution for an illustrative example of data
54: published by Fairall.\\
55: {\bf Keywords} Gamma distribution, void distribution, randomness, information geometry \\
56: MSC classes: 85A40; 60D05
57:
58: \end{abstract}
59:
60: \section*{Introduction}
61: A number of studies over the past ten years
62: have estimated the inter-galactic
63: void probability function and investigated its departure
64: from randomness. The basic random model is that arising from a Poisson
65: process of mean density $n$ galaxies per unit volume in a large box.
66: Then, in a {\em given region of volume} $V,$ the probability of finding
67: exactly $m$ galaxies is
68: \begin{equation}
69: P_m =\frac{(nV)^m}{m!} e^{-nV} \label{poisson}
70: \end{equation}
71: So the probability that the given region is devoid of galaxies is
72: $P_0=e^{-nV}.$ Then it follows that the probability density function
73: for the continuous random variable $V$ in the Poisson case is
74: \begin{equation}
75: p_{random}(V) = n \, e^{-nV} \label{negexp}
76: \end{equation}
77: For comparison with observations, the approximation fails for
78: very large $V$ since a finite volume box is involved in any catalogue.
79:
80: A hierachy of $N$-point correlation functions needed to represent
81: clustering of galaxies in a complete sense was devised by
82: White~\cite{white} and he provided explicit formulae, including
83: their continuous limit. In particular, he made a detailed study of
84: the probability that a sphere of radius $r$ is empty and showed
85: that formally it is symmetrically dependent on the whole hierarchy
86: of correlation functions. However, White concentrated his
87: applications on the case when the underlying galaxy distribution
88: was a Poisson process, the starting point for the present approach
89: which is concerned with geometrizing the parameter space of
90: departures from a Poisson process.
91:
92: \section*{Geometry of gamma models for void volume statistics}
93: We choose a family of parametric statistical models that
94: includes~(\ref{negexp}) as a special case. There are of course
95: many such families, but we take one that has been successful in
96: modelling void size distributions in terrestrial stochastic porous
97: media~\cite{dodsonsampson} and has been used in the representation
98: of clustering of galaxies~\cite{ijtp}. The family of gamma
99: distributions has event space $\Omega=\R^+,$ parameters
100: $\mu,\beta\in \R^+$ and probability density functions given by
101: \begin{equation}
102: f(V;\mu,\beta)=\left(\frac{\beta}{\mu}\right)^\beta \,
103: \frac{V^{\beta-1}}{\Gamma(\beta)}\, e^{-V\beta/\mu}
104: \label{feq}
105: \end{equation}
106: Then $\bar{V}=\mu$ and $Var(V)=\mu^2/\beta$ and we see that $\mu$ controls
107: the mean of the distribution while the spread and shape is controlled
108: by, $1/\beta,$ the square of the coefficient
109: of variation.
110:
111: The special case $\beta=1$ corresponds to the situation when $V$
112: represents the random or Poisson process in~(\ref{negexp}) with
113: $\mu=1/n.$ Thus, the family of gamma distributions can model a
114: range of stochastic processes corresponding to non-independent
115: `clumped' events, for $\beta<1,$ and dispersed events, for
116: $\beta>1,$ as well as the random case
117: (cf.~\cite{dodson,dodsonsampson}). Thus, if we think of this range
118: of processes as corresponding to the possible distributions of
119: centroids of extended objects such as galaxies that are initially
120: distributed according to a Poisson process with $\beta=1,$ then
121: the three possibilities are:
122: \begin{description}
123: \item[Chaotic or random structure] with no interactions among
124: constituents, $\beta=1;$
125: \item[Clustered structure] arising from mutually attractive
126: interactions, $\beta<1;$
127: \item[Dispersed structure] arising from mutually repulsive
128: interactions, $\beta>1.$
129: \end{description}
130: Figure~\ref{fgamma} shows a family of gamma distributions, all of
131: unit mean, with $\beta= 0.5, \ 1, \ 2.$
132:
133:
134:
135: \begin{figure}
136: \begin{picture}(300,220)(0,0)
137: \put(80,0){\resizebox{10cm}{!}{\includegraphics{fgamma}}}
138: \put(45,177){$f(V;1,\beta)$}
139: \put(125,160){$\beta=0.5$}
140: \put(180,70){$\beta=1$}
141: \put(310,50){$\beta=2$}
142: \put(300,-7){Void volume $V$}
143: \end{picture}
144: \caption{{\em Probability density functions, $f(V;\mu,\beta),$ for gamma
145: distributions of void volumes with unit
146: mean $\mu=1,$ and $\beta= 0.5, \ 1, \ 2.$ The case $\beta=1$ corresponds
147: to a `random' distribution from an underlying Poisson process.}}
148: \label{fgamma}
149: \end{figure}
150:
151:
152: Shannon's information theoretic `entropy' or `uncertainty' for such stochastic processes
153: (cf. eg. Jaynes~\cite{jaynes}) is given, up to a factor, by the
154: negative of the expectation of the logarithm of the probability density function~(\ref{feq}),
155: that is
156: \begin{eqnarray}
157: S_f(\mu,\beta) &=& -\int_0^\infty \log(f(V;\mu,\beta)\, f(V;\mu,\beta) \, dV \\
158: &=& \beta+(1-\beta)\frac{\Gamma'(\beta)}{\Gamma(\beta)} +
159: \log\frac{\mu \, \Gamma(\beta)}{\beta} \label{fentropy}
160: \end{eqnarray}
161: In particular, at unit mean, the maximum entropy (or maximum uncertainty)
162: occurs at $\beta=1,$ which is the random case, and then $S_f(\mu,1)=1+\log\mu.$
163:
164: The `maximum likelihood' estimates $\hat{\mu}, \hat{\beta}$ of $\mu,\beta$
165: can be expressed in terms of the mean and mean logarithm of a
166: set of independent observations $X=\{X_1,X_2,\ldots,X_n\}.$ These estimates are obtained
167: in terms of the properties of $X$ by maximizing the `log-likelihood' function
168: $$l_X(\mu,\beta)=\log lik_X(\mu,\beta)=\log\left(\prod_{i=1}^np(X_i;\mu,\beta)\right)$$
169: with the following result
170: \begin{eqnarray}
171: \hat{\mu}&=&\bar{X}=\frac{1}{n}\sum^n_{i=1}X_i\\ \log\hat{\beta}
172: -\psi(\hat{\beta})
173: & = & \overline{\log X} - \log\bar{X}
174: \end{eqnarray}
175: where $\overline{\log X}=\frac{1}{n}\sum^n_{i=1}\log X_i$ and
176: $\psi(\beta)=\frac{\Gamma'(\beta)}{\Gamma(\beta)}$ is the digamma
177: function, the logarithmic derivative of the gamma function
178:
179: The usual Riemannian information metric on the parameter
180: space $\cal{S}=\{(\mu,\beta)\in\R^+\times\R^+\}$
181: is given by
182: \begin{equation}
183: ds_{\cal{S}}^2=\frac{\beta}{\mu^2} \, d\mu^2 +
184: \left(\psi'(\beta) -\frac{1}{\beta}\right)\, d\beta^2 \ \
185: \ {\rm for} \ \mu,\beta\in\R^+ . \label{gammametric}
186: \end{equation}
187: For more details about the geometry
188: see~\cite{lauritzen,dodson,tg}. The 1-dimensional subspace
189: parametrized by $\beta=1$ corresponds to the available `random'
190: processes. A path through the parameter space $\cal{S}$ of gamma
191: models determines a curve
192: \begin{equation} c:[a,b]\rightarrow {\cal{S}}: t\mapsto
193: (c_1(t),c_2(t))\end{equation} with tangent vector
194: $\dot{c}(t)=(\dot{c}_1(t),\dot{c}_2(t))$ and norm $||\dot{c}||$
195: given via~(\ref{gammametric}) by
196: \begin{equation} ||\dot{c}(t)||^2=\frac{c_2(t)}{c_1(t)^2} \, \dot{c}_1(t)^2 +
197: \left(\psi'(c_2(t))
198: -\frac{1}{c_2(t)}\right)\,\dot{c}_2(t)^2.\end{equation}
199: The information length of the curve is \begin{equation}
200: L_c(a,b)=\int_a^b||\dot{c}(t)|| \, dt \end{equation} and the curve
201: corresponding to an underlying Poisson process has $c(t)=(t,1),$
202: so $t=\mu$ and $\beta=1=constant,$ and the information length is
203: $\log\frac{b}{a}.$
204:
205: As we know from elementary geometry, arc length is often difficult
206: to evaluate analytically because it contains the square root of
207: the sum of squares of derivatives. Accordingly, we sometimes use
208: the `energy' of the curve instead of length for comparison between
209: nearby curves. Energy is given by integrating the {\em square} of
210: the norm of $\dot{c}$
211: \begin{equation}
212: E_c(a,b)=\int_a^b||\dot{c}(t)||^2 \, dt. \label{energy}
213: \end{equation}
214: so in the case of the curve $c(t)=(t,1),$ the energy is
215: $\frac{b-a}{ab}.$ It is easily shown that a curve of constant
216: $\mu$ has $c(t)=(constant,t)$ where $t=\beta$ and
217: $\dot{c}(t)=(0,1);$ this has energy
218: $\log\frac{a}{b}+\psi'(b)-\psi'(a).$
219:
220: Locally, minimal paths joining nearby pairs of points in $\cal{S}$
221: are given by the autoparallel curves or geodesics~\cite{tg}
222: defined by~(\ref{gammametric}). Some typical sprays of geodesics
223: emanating from various points are provided in~\cite{ijtp}. The
224: Gaussian curvature of the surface $\cal{S}$~\cite{tg} actually
225: controls all of the geometry of geodesics and it is given by
226: \begin{equation}
227: K(\mu,\beta)= {\frac{\psi'(\beta) +\beta \psi''(\beta) }
228: {4(\beta\psi'(\beta)-1 }}\ \ \
229: \ {\rm for} \ \mu,\beta\in\R^+
230: \end{equation}
231: \begin{eqnarray}
232: K_{\cal{S}}(\mu,\beta)&\rightarrow& -\frac{1}{4} \ {\rm as} \
233: \beta\rightarrow 0\\ K_{\cal{S}}(\mu,\beta)&\rightarrow& -\frac{1}{2} \
234: {\rm as} \ \beta\rightarrow \infty
235: \end{eqnarray}
236:
237:
238: \section*{Void diameter statistics}
239: For a general account of large-scale structures in the universe,
240: see Fairall~\cite{fairall}.
241: Kauffmann and Fairall~\cite{kauffmannfairall} developed a catalogue
242: search algorithm for nearly spherical regions devoid of bright galaxies
243: and obtained a spectrum for diameters of significant voids. This had
244: a peak in the range 8-11 $h^{-1}Mpc,$ a long tail stretching at least to
245: 64 $h^{-1}Mpc,$ and is compatible with the recent extrapolation models
246: of Baccigalupi et al~\cite{bacc} which yield an upper bound on void diameters
247: of about 100 $h^{-1}Mpc.$ We shall return to the data of Kauffmann and Fairall
248: later in this section.
249:
250: Simulations of Sahni et al.~\cite{sahni} found strong correlation between
251: void sizes and primordial gravitational potential at void centres; void
252: topologies tended to simplify with time. Ghigna et al.~\cite{ghignabr} found in
253: their simulations that void statistics are sensitive to the passage from
254: CDM to CHDM models. This suggested that the void distribution is sensitive
255: to the type of dark matter but not to the transfer function between types.
256: CHDM simulations gave a void probability in excess of observations,
257: CDM simulations being somewhat better. Vogeley et al.~\cite{vogeley}
258: compared void statistics with CDM simulations of a range of cosmological
259: models; good agreement was achieved for samples of very bright galaxies
260: ($M<-19.2$) but for samples containing fainter galaxies the predicted
261: voids were reported to be `too empty'. Ghigna et al.~\cite{ghignabb} compared
262: observational data with Gaussian-initiated $N$-body simulations
263: in a $50 \, h^{-1} Mpc$ box and found that at the $2-8 \, h^{-1} Mpc$
264: scales the void probability for CHDM was significantly larger than observed.
265: Ghigna et al.~\cite{ghignabt} compared simulated galaxy samples with the
266: Perseus-Pisces redshift survey. The void probability function did
267: discriminate between DM and CDM models, the former giving
268: particularly good agreement with the survey.
269:
270: Little and Weinberg~\cite{little} used similar $N$-body simulations,
271: and found that the void probability was insensitive to the shape of the
272: initial power spectrum. Watson and Rowan-Robinson~\cite{watson}
273: found that standard CDM predictors do yield reasonably good void
274: probability function estimates whereas Voronoi foam models performed
275: less well.
276:
277: Lachi\'{e}ze-Rey and daCosta were of the opinion that the available
278: samples of galaxies were insufficiently representative, because of greater
279: apparent frequency of larger voids in the southern hemisphere.
280: Bernardeau~\cite{bernardeau} started from a Gaussian field and derived
281: an expression for the void probability function, obtaining
282: $$\frac{\log P_0}{nV} \approx -(nV \, \sigma^2)^{-3/7} \ \ {\rm for \
283: large} \ nV \, \sigma^2,$$
284: with $\sigma^2$ the variance of the number of galaxies in volume $V.$
285: This distribution has an extended large tail because it is asymptotically
286: more like the exponential of $-(nV)^\frac{4}{7}$ than the Poisson case
287: which decays like the exponential of $-nV.$
288: Cappi et al.~\cite{cappi} examined the dependence of the void probability
289: function on scale for a range of galaxy cluster samples, finding a general
290: scaling to occur up to void diameters of about $100 \, h^{-1} Mpc.$
291: Kerscher et al.~\cite{kerscher} used the void probability function to
292: obtain spatial statistics of clusters on scales $10-60 \, h^{-1} Mpc,$
293: obtaining satisfactory agreement in a model with a cosmological constant
294: and in a model with breaking of scale invariance of perturbations.
295:
296:
297: \begin{figure}
298: \begin{picture}(300,220)(0,0)
299: \put(80,0){\resizebox{10cm}{!}{\includegraphics{pdfD}}}
300: \put(43,160){$h(D;\mu,\beta)$}
301: \put(300,-7){Void diameter $D$}
302: \end{picture}
303: \caption{{\em Probability density function Equation~\protect(\ref{Dpdf}), for
304: distributions of void diameters with unit
305: mean, $\mu=1.244$ and $\beta=0.370.$ These parameter values are the best
306: fit for the data of Kauffmann and
307: Fairall~\protect\cite{kauffmannfairall}, also used in
308: Figure~\protect\ref{kfzcat}.}}
309: \label{hpdf}
310: \end{figure}
311:
312:
313: For our model we consider the diameter $D$ of a spherical void with
314: volume $V=\frac{\pi}{6} D^3$ having distribution~(\ref{feq}).
315: Something close to the random variable $D$ has direct representation in some
316: theoretical models, for example as polyhedral diameters in Voronoi
317: tesselations~\cite{weygaert,weygaerticke,coles}.
318:
319: \begin{figure}
320: \begin{picture}(300,220)(0,0)
321: %\put(90,-50){\special{psfile=histzcat.eps hscale=100 vscale=100}}
322: \put(50,0){\resizebox{12cm}{!}{\includegraphics{histzcat}}}
323: \end{picture}
324: \caption{{\em Histograms of fractions of space occupied by different classes
325: of void diameters, from Kauffmann and
326: Fairall~\protect\cite{kauffmannfairall} (right hand columns)
327: and predicted from the void diameter
328: distribution Equation~\protect(\ref{Dpdf}) (left hand columns)
329: fitted to the
330: same coefficient of variation and mean; the parameters found were
331: $\mu=1.244$ and $\beta=0.370.$ The class centres are in units of $200 \,km/s$
332: with a mean close to 7 units.}}
333: \label{kfzcat}
334: \end{figure}
335:
336:
337:
338:
339:
340: The probability density function for $D$ is given by
341: \begin{equation}
342: h(D;\mu,\beta)= f(\frac{\pi}{6} D^3;\mu,\beta)\, \frac{dV}{dD} = \frac{1}{\Gamma(\beta)}\,
343: \left(\frac{ \beta }{ \mu }\right)^\beta \,
344: \left(\frac{ \pi D^3}{6}\right)^{\beta-1} \, e^\frac{ -\pi\beta D^3}{6\mu}
345: \label{Dpdf}
346: \end{equation}
347: Then the mean $\bar{D},$ variance $Var(D)$ and coefficient
348: of variation $CV(D)$ of $D$ are given, respectively, by
349: \begin{eqnarray}
350: \bar{D}&=& \frac{
351: \Gamma(\beta+\frac{1}{3}) }
352: { \left(\frac{\pi\beta}{6\mu}\right)^\frac{1}{3} \Gamma(\beta) }
353: \label{meanD} \\
354: Var(D)&=& \frac{
355: \Gamma(\beta) \, \Gamma(\beta+\frac{2}{3}) -\Gamma(\beta+\frac{1}{3})^2 }
356: { \left(\frac{\pi\beta}{6\mu}\right)^\frac{2}{3} \Gamma(\beta)^2 }\label{varD} \\
357: CV(D)&=&\sqrt{
358: \frac{\Gamma(\beta) \, \Gamma(\beta+\frac{2}{3})}
359: {\Gamma(\beta+\frac{1}{3})^2}-1 } \label{cvD}
360: \end{eqnarray}
361:
362: The fact that the coefficient of variation~(\ref{cvD}) depends
363: only on $\beta$ gives a rapid fitting of data to~(\ref{Dpdf}).
364: Numerical fitting to~(\ref{cvD}) gives $\beta;$ this substituted
365: in~(\ref{meanD}) yields an estimate of $\mu$ to fit a given
366: observational mean. By way of illustration, this has been done in
367: Figure~\ref{kfzcat} for the ZCAT/SRC data from Kauffmann and
368: Fairall~\protect\cite{kauffmannfairall,fairall} Figure 8a (cf
369: also~\cite{fairall} Figure 6.5), both set to unit mean
370: diameter---the true mean for that catalogue was about $7\, h^{-1}
371: Mpc.$ The fitted values were $\mu=1.244$ and $\beta=0.370.$ This
372: fit is not particularly good if the reported peaks are not an
373: artifact but, qualitatively, we observe that the fitted value
374: $\beta=0.370$ is apparently significantly less than $1,$ which
375: would correspond to the random model. We conclude tentatively
376: that, for the ZCAT/SRC data subjected to the Kauffmann and Fairall
377: search algorithm, the new model suggests clumping rather than
378: dispersion in the underlying stochastic process. A {\em
379: Mathematica} NoteBook for performing the fitting procedure is
380: available from the author, who would welcome more sets of data.
381:
382: Suppose that the parameter space $\cal{S}$ of gamma-based models is
383: a meaningful representation of the evolutionary process
384: and that some coordinates such as $(\mu=1.244,$ $\beta=0.370)$
385: in $\cal{S}$ represent current data. Then geodesics in $\cal{S}$
386: through this point represent some kind of extremal path.
387: Moreover, we may consider a vector field $U$ on $\cal{S}$ such that
388: $(\mu=1.244,$ $\beta=0.370)$ is the present endpoint of an integral curve of
389: $U,$ the initial point of this curve being presumably in an epoch when
390: less clustering ($\beta>0.370$), a random state ($\beta=1$) or even dispersion
391: ($\beta>1$) was present at higher density.
392: It would be interesting to investigate the various candidate
393: cosmologies for their appropriate vector fields, via the statistics
394: of matter and voids they predict. The present gamma-related parametric statistical
395: models provide the means to convert the catalogue statistics into
396: the coordinate parameters and a background geometrization of the statistics
397: on which dynamical processes may be formulated.
398:
399: \section*{Acknowledgement} The author wishes to thank A.P. Fairall, F. Labini
400: and M.B. Ribeiro for helpful comments during the preparation of
401: this article.
402:
403:
404: \begin{thebibliography}{99}
405: \bibitem{bacc} C. Baccigalupi, L. Amendola and F. Occhionero.
406: Imprints of primordial voids on the cosmic microwave background
407: {\em Mon. Not. R. Astr. Soc.} 288, 2
408: (1997) 387-96.
409:
410: \bibitem{bernardeau} F. Bernardeau.
411: The gravity-induced quasi-Gaussian correlation hierarchy
412: {\em Astrophys. J.} 392, no.1, pt.1; 10 June 1992; p.1-14.
413:
414: \bibitem{cappi} A. Cappi, S. Maurogordato and M. Lachi\'{e}ze-Rey
415: A scaling law in the distribution of galaxy clusters.
416: {\em Astron. Astrophys.} 243, 1 (1991) 28-32.
417:
418:
419: \bibitem{coles} P. Coles. Understanding recent observations of the large-scale
420: structure of the universe. {\em Nature} 346 (1990) 446-.
421:
422: \bibitem{dodson} C.T.J. Dodson. Gamma manifolds and stochastic geometry. In:
423: {\bf Proceedings of the Workshop on Recent Topics in Differential Geometry}, Santiago
424: de Compostela 16-19 July 1997. {\em Public. Depto. Geometr\'{i}a y Topolog\'{i}a}
425: 89 (1998) 85-92.
426:
427: \bibitem{ijtp} C.T.J. Dodson. Spatial statistics
428: and information geometry for
429: parametric statistical models of galaxy clustering. {\em Int. J. Theor. Phys.}
430: 38, 10 (1999) 2585-2597.
431:
432: \bibitem{tg} C.T.J. Dodson and T. Poston. {\bf Tensor Geometry} Graduate
433: Texts in Mathematics 130, Second edition, Springer-Verlag, New
434: York, 1991.
435:
436: \bibitem{dodsonsampson} C.T.J. Dodson and W.W. Sampson. Modeling a class of
437: stochastic porous media. {\em Appl. Math. Lett.} 10, 2 (1997) 87-89.
438:
439: \bibitem{fairall} A.P. Fairall. {\bf Large-scale structure in the universe}
440: Wiley-Praxis, Chichester 1998.
441:
442: \bibitem{ghigna} S. Ghigna.Galaxy surveys and dark matter models: the statistics
443: of voids. In, {\bf Birth of the universe and fundamental
444: physics} Springer-Verlag, Berlin, Germany, 1995, 287-90.
445:
446: \bibitem{ghignabb} S. Ghigna, S. Borgani, S.A. Bonometto, L. Guzzo, A. Klypin,
447: J.R. Primack, R. Giovanelli and M.P. Haynes.
448: Sizes of voids as a test for dark matter models.
449: {\em Astrophys. J. Lett.} 437, 2, 2 (1994) L71-4.
450:
451:
452: \bibitem{ghignabr} S. Ghigna, S.A. Bonometto, J. Retzlaff, S. Gottlober
453: and G. Murante.
454: Void analysis as a test for dark matter composition?
455: {\em Astrophys. J.} 469, 1, 1 (1996) 40-7.
456:
457: \bibitem{ghignabt} S. Ghigna, S. Borgani, M. Tucci, S.A. Bonometto,
458: A. Klypin and J.R. Primack. Statistical tests for CHDM and Lambda CDM cosmologies. {\em Astrophys. J.} 479, 2, 1 (1997) 580-91.
459:
460: \bibitem{jaynes} E.T. Jaynes. Information theory and statistical inference.
461: {\em The Physical Review} 106 (1957) 620-630 and 108 (1957) 171-190.
462:
463: \bibitem{kauffmannfairall} G. Kauffmann and A.P. Fairall. Voids in the distribution of
464: galaxies: an assessment of their significance and derivation of a void spectrum.
465: {\em Mon. Not. R. Astr. Soc.} 248 (1990) 313-324.
466:
467: \bibitem{kerscher} M. Kerscher, J. Schmalzing, J. Retzlaff,
468: S. Borgani, T. Buchert, S. Gottlober, V. Muller, M. Plionis and H. Wagner.
469: Minkowski functionals of Abell/ACO clusters.
470: {\em Mon. Not. R. Astr. Soc.} 284, 1 (1997) 73-84.
471:
472: \bibitem{lachieze} M. Lachi\'{e}ze-Rey, L.N. Da-Costa and S. Maurogordato.
473: Void probability function in the Southern
474: Sky Redshift Survey. {\em Astrophys. J.} 399 (1992) 10-15.
475:
476: \bibitem{lauritzen} S.L. Lauritzen. Statistical Manifolds. In {\bf
477: Differential Geometry in Statistical Inference}, Institute of Mathematical
478: Statistics Lecture Notes, Volume 10, Berkeley 1987, pp 163-218.
479:
480: \bibitem{little} B. Little and D.H. Weinberg.
481: Cosmic voids and biased galaxy formation.
482: {\em Mon. Not. R. Astr. Soc.} 267, 3
483: (1994) 605-28.
484:
485:
486: \bibitem{piran} T. Piran, M.Lecar, D.S. Goldwirth, L. Nicolaci da Costa
487: and G.R. Blumenthal. Limits on the primordial fluctuation
488: spectrum: void sizes and anisotropy of the cosmic microwave background radiation. {\em Mon. Not. R. Astr. Soc.} 265, 3 (1993) 681-8.
489:
490: \bibitem{sahni} V. Sahni, B.S. Sathyaprakash and S.F. Shandarin.
491: The evolution of voids in the adhesion approximation
492: {\em Astrophys. J.} 431, 1, 1 (1994) p.20-40.
493:
494: \bibitem{vogeley} M.S. Vogeley, M.J. Geller, C. Park and J.P. Huchra.
495: Voids and constraints on nonlinear clustering of galaxies.
496: {\em Astron. J.} vol.108, 3; (1994) 745-58.
497:
498: \bibitem{watson} J.M. Watson and M. Rowan-Robinson.
499: The void probability function for flux-limited samples.
500: {\em Mon. Not. R. Astr. Soc.} 265, 4
501: (1993) 1027-34.
502:
503: \bibitem{weygaert} R. van der Weygaert. Quasi-periodicity in deep redshift surveys. {\em Mon. Not. R. Astr. Soc.} 249 (1991) 159-.
504:
505: \bibitem{weygaerticke} R. van der Weygaert and V. Icke. Fragmenting the
506: universe II. Voronoi vertices as Abell clusters. {\em Astron. Astrophys.} 213 (1989) 1-9.
507:
508: \bibitem{white} S.D.M. White. The hierarchy of correlation functions and its relation
509: to other measures of galaxy clustering. {\em Mon. Not. R. Astr. Soc.} 186,
510: (1979) 145-154.
511:
512: \end{thebibliography}
513: \end{document}
514: