UCINET 是一套全面的分析數據,及其他一維和二維的數據分析,可讀寫多種不同的格式化文字檔案,以及EXCEL的檔案。最高可處理32767節點。社會網絡分析方法包括核心措施、鑑定小組、分析角色、圖論基礎、排列型統計分析。另外包含強大的矩陣分析程式,例如矩陣代數、多元統計。
UCINET(University of California of Irvine Network Programms)由於其使用上較簡單,不論對初學者或專家同樣方便。UCINET如其他的網絡分析軟體,它處理的原始Data Set必須是由研究者所coding的actor-actor或actor-issue 矩陣資料,透過個人與個人或個人與事件間的"關係",電腦乃能辨識其處理的分析單位,並且透過不同的指令來作不同的分析,當然,如SPSS和SAS一般,UCINET原始資料的基本性質也決定了分析的層次,不過UCINET也有可能對資料作某種程度的轉型 (Transformation)。
Version 6.746-6.753
Changes
» Added new routine Network|Whole-network|Homophily/Influence|Continuous to measure overall network autocorrelation for continuous attributes. Essentially, it measures the correlation between having a tie, and having similar attribute values
» Added reciprocity routine to the CLI. Both arc and dyad reciprocity are computed, and for arc reciprocity, a comparison with density is also provided, as (in random graphs) the expected value of reciprocity is density. Syntax is reciprocity(
->dsp reciprocity(sampson)
» Added density comparison to the triplet transitivity routine in the CLI, as in a random graph, the expected value of triplet transitivity is density
» Added access to the DL Editor from the CLI. Syntax: DLEDIT [
->DLEDIT //opens the dl editor without reading in a file
->DLEDIT campnet //opens the dl editor and offers to open campnet in a variety of formats
» Improved transitivity calculation in both the menu and the CLI. It is now much faster. Also, for very large networks (5000 nodes and 5 million edges), it was giving the wrong answers due to overflow. Now it is correct.
» Added option in Tools|Testing Hypotheses|Dyadic|Logistic Regression to choose 1-tailed vs 2-tailed significance. Previously, only 1-tailed was available. Now, 2-tailed is the default
» internal changes only
Additions
» Data for Learning Ucinet.xlsx. This excel file, which is automatically installed with UCINET, has been updated
» New datasets Kracknet and Krackattr, and Karatenet and Karateattr added. They are just copies of krack-high-tec, high-tec-attributes, zackar and zachattr, respectively
» Fragmentation centrality. Added fragcent function to the CLI. It is somewhat faster than other fragmentation centrality routines in ucinet, especially for undirected data.
Syntax: fragcent(<dataset> [method:delete|isolate])
->mycent = fragcent(campnet)
The method parameter chooses between isolating the node whose centrality is being measured (keeping the total number of nodes the same), or deleting the node (which means we compare a fragmentation score for a network with N nodes with a fragmentation score for a network with N-1) nodes. The default is delete.
Fixes
» Data|Remove|Isolates. Fixed bug which caused it to treat missing values as ties. Fixed another bug which was causing it to print the original matrix, not the one with isolates removed.
» Two-mode centrality. The eigenvector routine in Network|Two-mode|Centrality was giving poor answers when the largest eigenvalue was repeated -- e.g., if the largest eigenvalue was not unique. However, even though the routine has been corrected to get the same results as R and Stata, we still recommend ignoring eigenvector centrality when the principal eigenvalue is repeated.
» SVD. The SVD procedure in the CLI was updated to match the 2-mode centrality routine discussed above
» Eigenvectors. The eigenvectors() function of the CLI was updated to give better answers in the case of repeated eigenvalues.
» Hubs and authorities. The Network|Centrality|Hubs and Authorities routine was updated to give better answers when the principal singular value is repeated. In addition, the singular values are now printed.
» Fixed Network|Dyadic measures|Reachability routine. It was failing to run with larger datasets. In addition, it now runs in 64-bit mode as well as 32-bit.
Requirements and Specifications
Windows operating system Vista or later. If you have a Mac or Linux, you can run UCINET via BootCamp, VMFusion Ware, Parallels or Wine.
The 32-bit version is the standard one and runs on both 32bit and 64bit Windows systems. A limited 64-bit version is available but does not have all UCINET functions
100mb of disk space for the program itself (not including your data)
The more RAM the better, but the 32-bit version can't take advantage of more than 3GB of memory. If you have large data and a 64-bit version of Windows, you can try experimental 64-bit version, in which case 8GB of RAM or more would be useful. Remember, however, that even if a really large dataset fits in memory, it may take too long to analyze.
While the absolute maximum network size is about 2 million nodes, in practice most UCINET procedures are too slow to run networks larger than about 5000 nodes. However, this varies depending on the specific analysis and the sparseness of the network. For example, degree centrality can be run on networks of tens of thousands of nodes, and most graph theoretic routines run faster when you have very few ties, no matter how many nodes you have.
It assumes that the software has been installed with the data in the folder C:\Program Files\Analytic Technologies\Ucinet 6\DataFiles and this has been left as the default directory.
When UCINET is started the following window appears.
The submenu buttons give access to all of the routines in UCINET and these are grouped into File, Data, Transform, Tools, Network, Visualize, Options and Help. Note that the buttons located below these are simply fast ways of calling routines in the submenus. The default directory given at the bottom is where UCINET picks up any data and stores any files (unless otherwise specified) this directory can be changed by clicking on the button to the right.
Running a routine
To run a UCINET routine we usually need to specify a UCINET dataset and give some parameters. Where possible UCINET selects some default parameters which the user can change if required. Note that UCINET comes with a number of standard datasets and these will be located in the default directory. When a routine has been run there is some textual output which appears on the screen and usually a UCINET datafile contain the results that again will be stored in the default directory.
We shall run the degree centrality routine to calculate the centralities of all the actors in a standard UCINET dataset called TARO. First we highlight Network>Centrality>Degree and then click
This will bring up a box as follows
If you click on the help button then a help screen will open which looks like this. The help file gives a detailed description of the routine, explains the parameters and describes the output that will appear in the log file and on the screen.