Note: To find revision information for really old versions of UCINET look in these two places:
- To see a list of known bugs and user wish list items, check the fix list
- To download the latest version, click download.
- Added new measures to Networks|Egonets|Longitudinal|Egonet Change.
- Added option to Tools|Dissimilarities|Std Vector. New option is average absolute difference.
- Also changed the way Euclidean distance and Manhattan distance were handling missing values. The old results were correct, but required looking in the manual to understand. The program was projecting what the distances would have been had there been no missing values.
- Added Burt's reinforced structural holes (RSH) measure to both menu and command line
- In menu, go to Networks|Ego Networks|Reinforced Structural Holes
- In command line, the syntax is
-><outfile> = rsh(<network> [NORM] [OUT|IN|INTERSECT|UNION])
->y = rsh(campnet norm union)
- the option NORM yields normalized outputs (divides by degree), and the option UNION defines the alters as all nodes that either have ties to or from ego
- Added iterative proportional fitting (IPF) to command line. Takes a matrix and iteratively adjusts values until all row sums are the same and all column sums are the same
- -><outfile> = ipf(<matrix> [DIAGOK])
- ->y = ipf(x diagok)
- the option diagok ensures that diagonal values are considered valued in 1-mode matrices. Otherwise, when the matrix is 1-mode, the default is to ignore diagonal values
- Added "transition" command to command line. Given a vector of values representing categorical states in chronological order, it counts the number of times that state A leads to state B, for all pairs of possible states. Sample usage:
- ->x = transition(states)
- x(i,j) has the number of times that state i was immediately followed by state j
- ->y = rowstochastic(x)
- this command creates a matrix y in which y(i,j) gives the probability of changing to state j given that you are currently in state i
- This version includes some internal test code that was causing installation problems. It should be fixed now, but if necessary, click on ucinetsetup.zip below to get version 6.516 instead.
- Changed Tools|Testing Hypotheses|Node level|Regression to add buttons for selecting variables (columns) by name instead of number. Just press the L button next to the field asking for the column number. The dropdown menu supports multiple selection.
- Added new function in the command line called AGGBY. It allows you to aggregate adjacency matrices based on a categorical attribute. So this is basically a facility to create density tables or image matrices as they are called in blockmodeling. For example, if you have ties (possibly valued) between individuals in an organization, and you have a vector indicating which department each person is in, you can create a new matrix of ties between departments. The syntax is:
-><outputfile> = aggby(<matrix> <vector> [SUM|AVG|MIN|MAX] [DIAGOK])
->tiesbygender = aggby(campnet gender)
->tiesbyrole = aggby(campnet col(campattr 2))
By default, the program sums the values in the input matrix, but by adding a keyword you can change this to an average, minimum or maximum. For square matrices the program automatically ignores the diagonal, so if this is not what you want you have to include the term DIAGOK as the last entry in the arguments.
- Added new function in the command line (matrix algebra) called CORRMAT. It allows you to correlate matrices, but doesn't bother with significance tests. The syntax is:
-><outputfile> = corrmat(<matrix> <matrix> ... [diagok])
->corrs = corrmat(samplk1 samplk2 samplk3)
For square matrices the program automatically ignores the diagonal, so if this is not what you want you have to include the term DIAGOK as the last entry in the arguments.
- Added new menu option Tools|Correlate Matrices Across Datasets which is similar to CORRMAT but can correlate matrix both within and between datasets. For example, if the input datasets are Campnet (which is a single matrix) and camp92 (which contains two valued matrices), the result will a 3-by-3 correlation matrix in which correlations among all three matrix are calculated.
- Fixed bug in Networks|Subgroups|Factions which was handling valued data oddly. Now it treats all values greater than 0 as a tie. If you want to treat values as values, use Tools|Cluster|Combinatorial instead
- Changed Networks|Centrality|Eigenvector so that have missing values down the diagonal of an input matrix are now automatically recoded to 0s
- Fixed bug in Data|Affiliations (2-mode to 1-mode) which occurred when using column mode on datasets containing multiple matrices. Was giving wrong answers.
- Updated the installation package to include additional helpfiles and latest version of netdraw
- Also added a few Alt-Backspace shortcuts
- Added additional option to Data|Affiliations (2-mode to 1-mode). New option is max of cross-minimums.