*************** How to use neutet.exe (Ver.20091028) *************** "neutet.exe" is a tool for calculation of item reference profiles using Dr. Shojima's Neural Test Theory (Shojima, 2008a). The usage is as follows. 0 Demonstration 1 Data File 2 Usage of this Program 2.1 Double Click (Interactive Mode) 2.2 Double Click (Program Mode) 2.3 Command 3 Setting File 3.1 Name of Files 3.2 Data 3.3 Parameters for Computation 4 Reference 5 History 6 License 0 Demonstration Try an analysis with a sample data. (1) Double clicke "neutet.exe". (2) You are asked "Is there a setting file? [y/n]". Then push "n" and return key. (3) Input data file name as "data.dat". (4) Input start column of response as "4". (5) Input number of items as "20". (6) Input number of latent ranks as "5". (7) Input file name of item reference profile as "result.csv". (8) When you are told to push any key, push any key and return key. (9) Item reference profiles are written into "result.csv". 1 Data File Write 0s, 1s or dots without a gap. "1" means correct, "0" means incorrect, and "." means missing. One examinee's data are written in one line. One item corresponds to one column. The following is an example of 5 examinees' response to 3 items. 100 101 110 111 111 2 Usage of this Program 2.1 Double Click (Interactive Mode) Double click "neutet.exe". You are asked "Is there a setting file? [y/n]". If you push "n" key and input a name of the setting file, the interactive mode starts. You are asked the following 5 questions. * data file name * start column of response * number of items * number of latent ranks * file name of item reference profile If you need more detailed information (for example, rank membership profile, and so on), prepare a setting file and use program mode. 2.2 Double Click (Program Mode) Double click "neutet.exe". You are asked "Is there a setting file? [y/n]". If you push "y" key and input a name of the setting file, the program mode starts. If you input an absolute path including spaces, don't use double quotations. 2.3 Command Input "neutet.exe" and a name of the setting file. For example, if a name of the file is "exercise.txt", execute the following command. neutet.exe exercise.txt If the absolute path includes spaces, write it between two double quotations. For example if a name of the file is "C:\Documents and Settings\hashimot\My Documents\exercise.txt", execute the following command. neutet.exe "C:\Documents and Settings\hashimot\My Documents\exercise.txt" 3 Setting File Write a setting as "variable = value" in each line. For example, if a name of the data file is "data.dat", write as follows. infile = data.dat Variable names have to be written in lower-case letters. Spaces and tabs are ignored. Phrases after #, %, * are ignored as comment. The following is what variables mean. 3.1 Name of Files A relative path starts from the folder where "neutet.exe" is. If a file name includes spaces, write it between two double quotations. e.g. infile = "C:\Documents and Settings\hashimot\My Documents\data.dat" * infile (required) data file name * outfile a file name of item reference profiles default: item reference profiles are not output. note: The profile of each item is written in each row. The first row is a line of labels which mean latent ranks. The first column is a sequence of item numbers. * rankfile a file name of examinees' latent ranks default: ranks are not output. note: The latent rank of each examinee is written in each row. If examinees' IDs are specified, the IDs are output in the first column and the ranks are output in the second column, and the first row is a line of labels. * rmpfile a file name of examinees' rank membership profiles default: rank membership profiles are not output. note: The profile of each examinee is written in each row. The first row is a line of labels which mean latent ranks. If examinees' IDs are specified, the IDs are output in the first column, and the first row is a line of labels. * worpfile a file name of weighted observation ratio profiles default: weighted observation ratio profiles are not output. note: The profile of each item is written in each row. The first row is a line of labels which mean latent ranks. The first column is a sequence of item numbers. * uorpfile a file name of unweighted observation ratio profiles default: unweighted observation ratio profiles are not output. note: The profile of each item is written in each row. The first row is a line of labels which mean latent ranks. The first column is a sequence of item numbers. * fixedfile a file name of fixed value of item reference profiles default: fixed value of item reference profiles are not input. note: The profile of each item has to be written in each row. The first row has to be a line of labels which mean latent ranks. The first column has to be a labels of item numbers. If you don't wish to fix a value of a certain rank at a certain item, write an asterisk (*) at that cell. When some items are not specified, the values of those items are not fixed. * dlm (value: comma / tab / space) delimiters of item reference profiles and examinees' latent ranks default: dlm = comma note: Even when it is written as "dlm = space", the format is not fixed. 3.2 Data * idbegin beginning column of each examinee's ID (counted from 1) default: An ID field does not exist. * idlength length of each examinee's ID column default: read until a space or a tab note: When "idbegin" is not written and "idlength" is written, the value of "idbegin" is "1". * databegin beginning column of response (counted from 1) default: the next column of an ID field (spaces and tabs are skipped) * nitems number of items defalut: read until the end of a line note: When you specifiy a variable "itemuse", you have to include ignored items into the value of "nitems". For example, if you use 10 items and ignore 5 items, nitems have to be set as 15. * itemuse whether each item is used or not default: all items are used note: Write 0s or 1s as many as the number of items. "1" means that this item is used, and "0" means that this item is ignored. For example, when you use the first 5 items and the last 10 items out of 20 items, write as follows. itemuse = 11111000001111111111 When some items are ignored (written as "0"), all responses to those items are treated as missing. So if you'd like to fix item reference profiles of some items, you have to set the values of those items as "1"s. * firstobs the first observation used in calculation default: firstobs = 1 * nobs the number of observations used in calculation default: observations to the end of the file are read. 3.3 Parameters for Computation * nranks number of latent ranks default: nranks = 10 * method (value: ls / ml / gtm) winner determination method default: method = ml note: "ls" means square of Euclidean distance. "ml" means most likelihood method. "gtm" means Shojima's method using generative topographic mapping (Shojima, 2008b). This parameter affects rank membership profiles. * iter number of iteration default: iter = 100 note: You can write this variable as "maxiter". * increase (value: true / false) monotone increase constraint of each item reference profile default: increase = false * seed random number seed for shuffling data default: automatically generated from the current time * alpha1, alphat, sigma1, sigmat $\alpha_1$, $\alpha_T$, $\sigma_1$, $\sigma_T$ in Shojima (2008a) default: alpha1 = 1.0; alphat = 0.1; sigma1 = 1.0, sigmat = 0.12 note: A fraction is not allowed as a value. e.g. alpha1 = 0.333 (OK) alpha1 = 1/3 (NG) Values are not checked. * phi $phi$ in Shojima (2008b) default: See equation (21) in Shojima (2008b). When number of ranks are more than 20, phi = 1/3. note: This is an option for generative topographic mapping. If you write as "phi = 1", then smoothing matrix is not used. 4 Reference Shojima, K. (2008a). Neural test theory: a latent rank theory for analyzing test data. DNC Research Note, RN-08-01. Shojima, K. (2008b). The batch-type neural test model: a latent rank model with the mechanism of generative topographic maping. DNC Research Note, RN-08-06. 5 History * 2007/Sep/10 Released. * 2007/Sep/11 The variables "rankfile", "crit", "diffcrit" have been added. * 2007/Oct/02 Default value of alpha1 has been changed from 1/Q to 0.1. * 2007/Oct/04 The variables "rmpfile" and "method" have been added. * 2007/Oct/15 The variables "worpfile" and "uorpfile" have been added. * 2007/Nov/21 The variable "fixedfile" have been added. * 2007/Dec/17 Default value of maxiter has been changed from 100 to 500. Default value of sigma0 has been changed from 1.0 to 1.2. * 2008/Aug/22 Learning parameters have been updated to those of Shojima (2008a). The variables for stopping criteria ("crit" and "diffcrit") have been deleted. The variable "maxiter" has been changed to "iter", and default value has been changed from 500 to 100. However, you can use also "maxiter". The variables "itemuse", "firstobs", "nobs" have been added. Interactive mode has been added. * 2008/Sep/10 Character checking has been added. * 2009/Oct/28 Computation program of generative topographic mapping (Shojima, 2008b) has been added. The variable "phi" has been added. 6 License Copyright (c) 2007--2008 HASHIMOTO Takamitsu and SHOJIMA Kojiro. All rights reserved. THIS SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY. IN NO EVENT SHALL THE COPYRIGHT HOLDERS BE LIABLE FOR ANY DAMAGES ARISING FROM THE USE OF THIS SOFTWARE.