ANU-MACRONET2 . A PROGRAM FOR COMPUTING SYMMETRIC ADJACENCY, REACHABILITY, DISTANCE AND POINT-CENTRALITYIN LARGE NETWORKS (N< 3000)D .W . Carrick, (Sociology, Research School, A .N .U .) <strong>and</strong> R .A . 0modei <strong>and</strong> A .S . KZovdahZ (Sociology, AustralianNational U .)ANU-MACRONET2 is an improvement on the previous version of MACRONET (Klovdahl <strong>and</strong> Omodei, 1978) . Ituses a new algorithm that makes it possible to compute reachability, distance <strong>and</strong> point-centrality (Freeman,1977) in networks with up to about 3,000 nodes (Carrick, Omodei, <strong>and</strong> Klovdahl, 1978) . The exact upperlimits of the program depend on the amount of core space available, <strong>and</strong> on the number of links relativeto the number of nodes in a network .For example, using 48K words of core on the ANU's UNIVAC 1100/42 a network of about 600 nodes <strong>and</strong>36,000 links (i .e . an average of 60 links per node) would take approximately 9 minutes (UNIVAC SUP time)to process ; a network of about 2,000 nodes <strong>and</strong> 20,000 links (10 per node) would take on the order of 22minutes ; <strong>and</strong> a network of about 2,800 nodes <strong>and</strong> 14,000 links (5 per node) would require about 24 minutes .(A procedure for estimating the time required for processing a particular network is provided in a moredetailed description of the program (Carrick, Omodei, <strong>and</strong> Klovdahl, 1978) .In this version of MACRONET there is also provision for a variety of input formats <strong>and</strong> output specifications. Measures of centrality, as noted, may also be obtained as part of the output .ANU-MACRONET2 is written in FORTRAN V .References :Carrick, D .W ., R .A . Omodei, <strong>and</strong> A .S . Klovdahl . ANU-MACRONET2 : A program for analyzing data from largenetworks (n< 3000) . Computer program description . Department of Sociology, Faculty of Arts, TheAustralian National University, Canberra, 1978 .Freeman, L .C . A set of measures of centrality based on betweeness . SOCIOMETRY, 1977, 40(1), 35-41 .Klovdahl, A .S . <strong>and</strong> R .A . Omodei . ANU-MACRONET : A set of programs for computing symmetric adjacency, reachability,<strong>and</strong> distance matrices for larger networks (n< 60) . CONNECTIONS, 1978, 1(2), 19 .ANU-R4NDNET . A PROGRAM FOR GENERATING MONTE CARLO SAMPLES OF NETWORKS .Alden S . KZovdahZ (Sociology, Arts, Australian National U .)This program makes it possible to take r<strong>and</strong>om samples from a constructed (population) network . Twobasic types of sampling are possible : "Independent r<strong>and</strong>om sampling", <strong>and</strong> "dependent r<strong>and</strong>om sampling" .In the former, nodes are selected independently of links that might connect them to other nodes . In thelatter, which may also be referred to as link-tracing, the selection of nodes (other than the first in asequence) is dependent on the linkages actually observed in a population .Once the sample selections have been made, by either method, the resulting data is converted intomatrix form for analysis using such network programs as MICRONET, MACRONET2, <strong>and</strong>/or MARKOV (Klovdahl <strong>and</strong>Omodei, 1978 ; Carrick, Omodei <strong>and</strong> Klovdahl, 1978 ; <strong>and</strong>, Thong <strong>and</strong> Klovdahl, 1978) .Ther results of a large number of samplings, of course, may then be used to evaluate the relativemerits of different approaches to network sampling .ANU-RANDET is written in FORTRAN VReferences :Carrick, D .W ., R .A . Omodei <strong>and</strong> A .S . Klovdahl . ANU-MACRONET2 : A program for computing symmetric adjacency,reachability, distance <strong>and</strong> point-centrality in large networks (n< 3000) . Computer program description .Department of Sociology, Faculty of Arts, The Australian National University, Canberra, 1978 .Klovdahl, A .S . <strong>and</strong> R .A . Omodei, ANU-MICRONET : A set of programs for computing symmetric adjacency, reachability,symmetric reachability <strong>and</strong> distance matrices for smaller networks (n< 85) . CONNECTIONS, 1978,1(2),19 .Thong, L .D . <strong>and</strong> A .S . Klovdahl, ANU-MARKOV : An interactive program for using Markov statistics to analyzenetwork data . Computer Program Description . Department of Sociology, Faculty of Arts, The AustralianNational University, Canberra, 1978 .
ANU-MARKOV . AN INTERACTIVE PROGRAM FOR USING MARKOV STATISTICS TO ANALYZE NETWORK DATA .L .D . Thong <strong>and</strong> A .S . KZovdahZ (Sociology, Arts, Australian National U .)Markov methods provide the means to answer important questions about networks (<strong>and</strong> other social phenomena). But, although these methods have previously been used to make inferences about networks (e .g . White,1959 ; Beshers <strong>and</strong> Laumann, 1967 ; Klovdahl, 1976), to date there has been little systematic effort to developthe programs (or packages) that would make Markov methods more readily accessible for network analysis .ANU-MARKOV is an attempt to facilitate the use of Markov statistics for analyzing network data . Thisprogram computes maximum likelihood estimates of transition probabilities from raw data, <strong>and</strong> then allowsfive relevant statistical tests to be carried out . More specifically, data believed to result from aMarkov process can be tested (a) for the specific values of the transition probabilities, (b) for stationarity,(c) for order <strong>and</strong> (d) for homogeneity (i .e . to ascertain whether or not data from multiple samples weregenerated by the same process) . In addition, there is a test for the possible relationship between twosets of different states . These tests were developed by Anderson <strong>and</strong> Goodman (1957) . Finally, the appropriatecorrections for (finite) sample size are also calculated (Sharp, 1975) .The limits of the program depend on the particular test involved . At present, chains of up to 11 statesmay be analyzed, the maximum order that can be tested is 3, <strong>and</strong> up to 50 transitions may be processed (inthe stationarity test) .The program is written in FORTRAN V .References :Anderson, T .W . <strong>and</strong> L .A . Goodman . Statistical inference about markov chains . ANNALS OF MATHEMATICAL STATISTICS, 1957, 28, 89-110 .Klovdahl, A .S . <strong>Networks</strong> in society : The r<strong>and</strong>om walk strategy . Working paper . Department of Sociology,Faculty of Arts, The Australian National University, Canberra, 1976 .Sharp, S .A . Corrections to likelihood ratio tests of hypotheses concerning the parameters of Markov chains .BIOMETRIKA, 1975, 62(3), 595-598 .White, H .C . Search parameters for the small world problem . SOCIAL FORCES, 1970, 49, 259-264 .SNAP -SOCIAL NETWORKS ANALYSIS PACKAGEClive Payne, David Deans <strong>and</strong> Clyde Mitchell (Nuffield College, Oxford)Greg Heil provided a useful review of the final specification for SNAP in <strong>Connections</strong> Vol II - No . 1but we would like to point out that there appears to be a slight misunderst<strong>and</strong>ing . As yet the package hasnot been implemented . The specification represents the results of only the first stage of the project<strong>and</strong> we can only proceed to the second stage of full implementation if we can demonstrate to our sponsorsthat there is sufficient dem<strong>and</strong> for the package . Accordingly readers of <strong>Connections</strong> are cordially invitedto send for a copy of the final specification . We would be most grateful to receive indications of supportfor the further work on the package . Furthermore, this way of proceeding with package development givespotential users the opportunity to suggest extensions <strong>and</strong> improvements to be incorporated in the finalversion of the package .