What MultiNet does

MultiNet does ordinary data analysis of ordinary data -- univariate descriptive statistics (mean, median, mode, standard deviation, frequencies, cumulative frequencies, etc.), and it makes this kind of analysis very easy to do. But there are a lot of packages that do this. What makes MultiNet different?

It's different from just about anything you'll see in four ways.

First, it supports the user to an unusual extent. Everything it does is shown both graphically and in ordinary tabular form. It helps you do the kinds of things that need to be done when you're working with complicated data -- recode variables, create new variables, transpose a set of variables and values (so the values become variables and the variables become values), combine sets of ordinary variables into meta-variables so you can examine a set of relationships in parallel, and more.

Second, it combines data that describes people with data that describes relationships between people into a single analytic model. It lets you mix node variables with link variables in a variety of kinds of analysis to explore the patterns in your network. While most network programs perform one or another type of structural analysis, MultiNet does contextual analysi: it looks at attributes of people in the context of the relationships between and among them, and it looks at charactistics of relationships between people in the context of the attributes of the people.

Third, it is very happy with both ego-centric and ordinary whole-network data. It can easily deal with data that has many variables describing attributes of nodes and many that describe relationships between nodes.

Fourth, it can do four kinds of spectral analysis (eigen decomposition). With any eigen analysis you can see one-, two-, or three-dimensional projections of the results on the screen. You can rotate the three-dimensional results about any axis. You can zoom in or out. You can create new node variables from the eigenvectors. You can have the program partition the network on the basis of the eigenvectors to locate clusters or structurally equivalent nodes.

MultiNet started as an extended replacement for FATCAT, but it does so much more that there is little in common beyond the data model it uses. MultiNet is an interactive menu-driven program for exploratory analysis and display of discrete and continuous multivariate network data. It has context-sensitive, interactive, on-line help, and always presents a color graphic representation of the data or the results of analysis. All graphics can be saved as bitmap or PostScript files. The program does ordinary univariate descriptive statistics, crosstabulation, analysis of variance, regression, and correlation. It also does network versions of crosstabulation, anova, correlation-regression in which it combines data that describes nodes with data that describes relationships between nodes into a single analytic model. It lets you mix node variables with link variables in a variety of kinds of analysis to explore the patterns in your network. While most network programs perform one or another type of structural analysis, MultiNet also does contextual analysi: it looks at attributes of people in the context of the relationships between and among them, and it looks at charactistics of relationships between people in the context of the attributes of the people. It is very happy with both ego-centric and ordinary whole-network data. It can easily deal with data that has many variables describing attributes of nodes and many that describe relationships between nodes.

The program has a variety of flexible data manipulation capabilities. It can handle missing data. It performs continuous and discrete transformations, such as ordination, quantiles, recategorization. Sets of ranked variables can be inverted. It does linear, log, power, and z transforms. New variables can be created by transforming or combining existing ones in any manner describable by algebraic equations. The program also provides file viewing and editing capabilities. It can do four types of eigen decomposition of networks with up to 5,000 nodes for spectral analysis with interactive graphical display of results in 1, 2, or 3 dimensions, including link direction and/or strength, node attribute labels, and more options for graphic representation of eigen analysis results. The results of eigen analysis are integrated with the rest of the program so coordinates in eigen space can be used as variables in any other analysis the program does. Results of eigen decomposition can be used to create partitions that identify clusters or sets of structurally equivalent nodes. MultiNet does p* analysis on networks with up to 5,000 nodes, with interactive graphical display of results. Results of eigen analysis can be used to improve p* fits when using block structures. There is no easier way to do eigen analysis and p* modeling of networks.

Go back.