Here is a list of the files in the MultiNet archive and a brief description of what is in each of them: 201.lin & 201.nod -- a pair of ASCII files that describe the students in a course and the 16 people most important to each student and the type of contact the students have with them. This is an go-centric network. These files can be "imported" into MultiNet. 301.lin & 301.nod -- a pair of ASCII files that describe the students in another course and all of the interactions each of them had with anyone else over a 1-week period. This is an ego-centric network. These files can be imported into MultiNet. kids.lin & kids.nod -- a pair of ASCII files that describe the kids in a daycare center and who they were seen playing with and who they say they play with. This is not an ego-centric network; there is data for all relationships between network members. These files can be imported into MultiNet. kids.mnw \ These are MultiNet system files created by "Saving" the 201.mnw } data in the above pairs of files. (These mnw files can 301.mnw / be"loaded" into MultiNet. "Save" and "Load" are MultiNet commands, accessible from the "File" menu in MultiNet.) nodkids.xls & linkids.xls -- these are spreadsheet files with the kids link and node data (these files can be opened in MS excel or Corel Quattro.) nodkids.csv & linkids.csv -- comma-separated-value (CSV) files produced by saving the above spreadsheet data as "CSV" These can be imported into MultiNet. adj2neg.exe -- a program that converts adjacency matrices into link-list data of the type that MultiNet and Negopy use. free2fix.exe -- a program that reads a file in which values are separated by some non-numeric character (commas, tabs, letters, dollar signs, etc....) and produces files that can be read by MultiNet. wmn.ico -- an icon file for MultiNet c23220.dll, lanc32.dll, pspar32.dll -- parts of MultiNet aplwr.adf, aplwr.exe -- parts of the apl system that runs MultiNet mnw.mni -- initialization file for MultiNet WMN4_74R.w3 -- the APL workspace for MultiNet START_MN.bat -- the DOS bat file that you double-click on to start MultiNet WMNHELP.txt -- the text contents of the on-line documentation and help system for MultiNet ------------------------------------------------------------------------ ASCII data files (such as kids.nod and kids.lin) must have names that meet these criteria: 1) the part of the names that come before the "." must be the same (eg. "kids"); 2) the part of the name of the file that describes nodes must end in ".nod" (eg. "kids.nod"); and 3) the part of the name of the file that describes links must end in ".lin" (eg. "kids.lin") To create names for CSV files (Comma separated value) that will be recognized by MultiNet, do the following: 1) start with a single word that gives a general name to your data (eg. "kids"); 2) for the file that describes nodes, combine "nod" and the word from the first step (eg. "nod" + "kids" = "nodkids"); 3) for the file that describes links, combine "lin" and the word from the first step (eg. "lin" + "kids" = "linkids"); 4) add ".csv" to the end of the result of step 2 and 2 (eg. "nodkids.csv" and "linkids.csv") ------------------------------------------------------------------------- You might be interested in some of our research papers which discuss issues at the core of MultiNet. You can find them if you go to http://www.sfu.ca/~richards and click on "Research Papers" on the left near the top of the screen. MultiNet is shareware. If you really like MultiNet and find it useful, please consider the possibility of sending a check for $20 written to Andrew Seary. He is my graduate student and programmer, and I wish to give him some support when I can. Please send the check to Andrew at 10 Boundary, Burnaby, BC, Canada V5K 4R3. (If you have already done this, there is no need to do it again unless you have an excess of $$ and are looking for a happy home for some of it.) Here are some brief instructions to use Multinet. This will give you some idea how the program works. (There is a lot more in the manual.) Double-click on START_MN.bat to make MultiNet start. To make MultiNet stop, click on the little "x" button on the upper right corner of the program's window.. Click on "File" to import or load data files. Choose "Load" to get access to data files that have been used in MultiNet and saved as system files. Select "Kids.mnw" to get the kids daycare data. (This data was collected by students in a course I taught a few years ago. They watched kids in a daycare centre and made notes of who they saw playing together and they asked the kids who they played with.) ASCII data files must be in right-justified fixed-format. This means that each variable has a specified number of characters associated with it in each line of data. The ID numbers might be 3 digits wide. The variable "sex" might be 2 digits wide. This will give you one blank space and a "1" or a "2". Age might be three digits wide -- a blank space followed by a two-digit number. If you have many people who are over 99 years old, use 4 digits for age. When you print your data file using a monospaced font like Courier, the numbers line up in neat columns because a variable occupies the same number of spaces on each line of data. To import "standard" ASCII cases by variables data files (these are not network data files), click on "File" then "Import" and then "Standard" and then select the file you want. Standard data files for MultiNet must have the extension "dat". To import ASCII network data files, click on "File" then "Import" and then "Network" and then select the file you want. For ASCII network data, you must always have a pair of files and the one that describes nodes has the extension "nod" while the one that describes links has the extension "lin". Other than the differences in extensions, the file names must be the same. For example you might have mydata.nod and mdata.lin (but not mydata1.nod and mydata2.lin). You *may* see one irritating problem: the two windows which show you the variables in the two data files may be on top of each other. Drag them around in the same way you drag other windows. Click on "Eigenspaces" on the menu bar. Then click on "Link". Now choose either "say" or "play" (who they say they play with or who they were seen playing with). Then click "OK". Now choose a method of eigen analysis. I suggest you pick "Normal." Now move the little "Explore" window to the side. Now click on "Node" on the menu bar and select "sex". This will make blue dots for the boys and red for the girls. Now click on the "+Y" or "+X" buttons on the "Explore" window. Right-click on the "+Z" to zoom in and the "-Z" to zoom out. Click on "Quit" on the menu bar to do a different analysis. Click on "Analyse" then "Network". Then click on "Type" on the menu bar and then "Xtabs". Now select the rows variable by clicking on "Rows" Select "From Nodes" and then "Sex" to select the sex of the kids from whom links come. Now select the columns variable clicking on "Cols" Select "To Nodes" and then "Sex" to select the sex of the kids to whom links go. Now click "Go" on the menu bar and then "Link" and then "play" then "OK". Type anything you want for a descriptive label. Now you see a "panigram" which is a graphical representation of a crosstabs table. Click "Help" and "Interactive". Use the arrow buttons on the XTAB Help window and see what happens. Read the text on the bottom of the graphic when you do this. Want to see the tabular report? Close the XTAB Help window. Then click on "Report" on the menu bar and "View". Have fun! With MultiNet you can do spectral analysis on networks with up to 5,000 nodes. If your networks are bigger than that and you need spectral analysis, let us know. We can increase the size. Some documentation is available in "Help." Three sample datasets are in the zip file: kids, 301, Spss. The first two are network data; the third is a standard rectangular cases by variables data file (the sample data that came with a version of PC SPSS). Network datasets each have two files; one describes the nodes (filename.nod) and the other describes the links (filename.lin). Standard datasets have only one file (filename.dat). You can "load" files saved as MultiNet system files -- kids.mnw or 301.mnw; and you can "import" ASCII files, either network file-pairs or single standard data files. To "import" kids.nod and kids.lin, you tell the program to import kids.nod and it will automatically import the other file -- kids.lin. The file extensions have these meanings: nod = node variables; lin = link variables; dat = "standard" (cases X variables) file. p* is accessible through the "Model" menu item. The program makes a graphic display for p*. You can do p* on networks with up to 5,000 nodes. There are a few minor irritating things we are still working on, but they have to do with where certain windows appear on the screen and stuff like that. ------------ About the format of files for MultiNet ... MultiNet 4.75 can read tab- or comma-delimited files (.CSV) in addition to the fixed-format described below. Below is a part of the beginning of a data file you would use for MultiNet. This example is for when you have a single cases-by-variables file that you want to examine. [If you want to do a network analysis, you need a pair of files, one which describes the people and one which describes the connections between them. The formats of these files are the same as the one described below with one exception: the file that describes connections between people must have two ID variables at the beginning of each line. The first one must be called "ID1" and the second must be called "ID2". The first is the ID number of the person the link is from and the second is the ID number of the person the link is to. The file that describes the nodes must have the extension ".nod" and the file that describes links must have the extension ".lin" and both should have the same name except for the extensions. The files should go in the same folder that the rest of the MultiNet files are in.] Now on with the description of the sample file below.... The data in the sample file here describes people's assessments of a number of events that might happen. They were asked how likely it is that the event will happen, when the event will happen, and how important the event is likely to be. They were asked these questions for about twenty events. The first line of the file contains "ID (1-6)". This means that the name of the first variable is "ID" and the variable is the first 6 characters on a line of data. The second line "Prob1 (8)" says that the name of the second variable is "Prob1" and the variable is the 8th character on a line of data. The third line is a "/" which means that you are going to define value labels for the variable Prob1. The next five lines contain the value label definitions. For the value "1" the label is "highly improb". For the value "2" the label is "improbable".etc. The line after the fifth value label contains only a "/" and this signifies the end of value labels for the variable Prob1. The next line says the name of the third variable is "Import1" and it appears in the 9th character of a line of data. And so on. The word "end" appears on the line between the description of the last variable and thefirst line of data. The value labels are optional. I recommend that you use them. I recommend that you make the actual labels no more than 12 characters long if that is possible. The variables must appear in the list at the beginning of your data file in the same order that they appear in a line of data. If you list the variables out of order, the program will choke and complain. If you are using commas to separate the answers for one question from the answers to the next question (as I did in the example below) the commas are not part of the actual variables and the characters they occupy should not be counted in the location of the variables. For example, there is a comma in the 7th character of a line of data. This is not part of the ID variable or part of the Prob1 variable. This is why ID appears in "(1-6)" and Import1 appears in "(8)" and the 7th character isn't even mentioned. Since commas will be the 7th, 11th, 15th, 19th, ... characters on each line of data, these character numbers do not appear in the location of any variables. All variables (except IDLABEL which may be used to supply names for nodes, and which must appear immediately after the ID variable) must have numeric values. No letters or punctuation marks unless a particular one is used for the missing value code. I recommend that you use "^" for the missing value code because you can see it if you print the file or open it for editing. When you have a single cases-by-variables file like this example, the file's name must end in ".dat" When you are using MultiNet, to work with a file like this, you are working with a "standard" file and will do "standard" analysis. When you have a pair of files (one for people and one for connections between people), you have "network" data and can do "network" or "standard" analyses. To save a file containing one of the graphics you see on the screen, click on "Graphics" and then "Bitmap". To do univariate frequency distributions, click on "Variables" then "Select" then "Node" and then choose a variable. To see the next variable click on "Next". When you are done with univariate frequency distributions, click on "Quit" which will take you back to the main menu screen. Now you can choose "Analyse" to do an analysis. I think you will understand what to do if you look at the menu items that are available and not greyed out. You can always click on "Help" to get help. In the example file below, ^ is the missing data character. ----here is the beginning of the data file------------------------ ID (1-6) Prob1 (8) / 1 highly improb 2 improbable 3 no opinion 4 probable 5 highly prob / Import1 (9) / 1 not import 2 little import 3 no opinion 4 import 5 highly impt / Time1 (10) / 1 pre-2005 2 2005-09 3 2010-15 4 post-2015 5 no opinion / Prob2 (12) Import2 (13) Time2 (14) Prob3 (16) Import3 (17) Time3 (18) : : end 00001,431,234,223,123,112,^^^,121,153,^^^,^^^,^^^,... 00002,131,122,143,112,431,145,244,112,343,554,112,... ------------ Here's another little exercise to familiarise yourself. -- From the main menu choose "File" then "Load" and select "kids.mnw" You will get 2 windows showing all the Node (attribute) and Link (network) variables. These are for information only. -- From the main menu choose "Analyse" then "Network". Then choose "Type" and then "XTABS". -- From the menu bar choose "Rows" then "From Node" and from the list of Node variables that appears, choose "SEX" -- From the menu choose "Cols" then "TO Node" and then choose "SEX" again. Choose "GO" then "link" and then choose "SAY" from the list of link variables. You have now defined a network analysis which will sort actors by SEX (boy and girls), where the network is who each of them SAY they play with. You will be asked for a comment (which goes on the display). Enter a comment and press "OK" or just press "OK". The display shows you who SAYS they play with whom, sorted by SEX. Press "Report" then "View" for details of the crosstab tables. Scroll down the report until you see the value for Chi-squared. Press "Quit" to leave the report, and "Quit" to leave the Analysis. Press "Eigenspaces", then press "Link" and select "SAY" again. Then select "CorrAnal". A picture will appear of the SAY network. You can spin the network around on any axis with the Explore window. Press "Node" and select "SEX". Now the Nodes of the network are coloured by sex. Click "Show" in the little Explore window and select "Labels". Now you see the names of the children in the network. Now select "Dimensions" then "1D". You will get a different display, like an adjacency matrix, with coordinates coming from the first eigenvector. The diagonal is coloured by SEX (since you already selected this). Notice the 2 clusters, and the relationship to sex. Press "Define" then "Partition" then "OK" and "OK". This creates a new Node variable with attributes -1 and +1, which determine which cluster each actor is in, based on the signs of their values on the first eigenvector. Press "Quit" to get back to the main menu. Repeat the "Analysis" then "Network" and "Type" then "XTABS using the new variable (automatically called 2PC.SAY, though you can over-ride this.) Look at Chi-squared in the report. It's bigger than the one you saw earlier. The eigenevector signs do a better job of sorting who SAYs they play with whom than sex. Now get back to the main menu and press "Models" then "pstar". Select "Link" and "SAY" then "Node" and "SEX". Then press "Pstar" and select "1" and "2" (click the mouse to select "1", then shift+click to select "2"). Press "OK" for the blocking and parameters. Now you see a picture showing the p* fit to the SAY network, based on blocking by SEX. In the Explore window, select "Show" then "Node Perm". This shows the network permuted by the attributes of SEX. Note the value of -2*LogPL on the left. It's a measure of "badness of fit": the bigger, the worse. Repeat the Pstar fit with "Node" - "2PN.SEX". You will see a smaller -2*LogPL, meaning the fit is better. The table underneath also shows more links have been predicted correctly by p*. That's an overview on how to use MultiNet. There's lots more, but this should get you started. Enjoy!