Quick Start Guide
The Web page, Recipe_sim.htm, contains a JavaScript program that simulates data for a Recipe design according to the relative-weight averaging model. This program is designed to be compatible with Recipe_Wiz.htm, and it produces data that can be analyzed in the same way as empirical data generated by the forms created by Recipe_Wiz.htm.
You can adjust parameters to explore predictions of the data. The simulation program also makes explicit exactly how the model functions. All of the code is contained inside this file.
To use it for the first time, change the title and short title and accept the defaults that are in the fields.
You can enter names for Factors A, B, and C, the weights of the factors (wA, wB, wC), the numbers of levels of each factor(nA, nB, nC) [in the JavaScript these are r, c, and l], the scale values of the factors (ai, bj, ck), and the weight and scale value of the initial impression (w0, s0). In the program, the scale values of factors A, B, and C are in arrays labeled row(k}, col(j), and lev(m).
It is assumed that responses in this task are a linear function of the subjective impressions given by the averaging model. You can change the values of the linear coefficients in the appropriate fields above.
Computer generated, "random" error can be added to each judgment. The program implements errors that are normally distributed with a mean of zero, and the standard deviation of the normal distribution can be specified by the user. For perfect predictions, set the standard deviation to zero (0). You need only simulate one "subject".
Finally, the user can also specify the precision with which the simulated responses should be rounded (e.g., 7.62, 7.6, or 8), which will tend to add an additional granularity to the data. If you plan to plot the predictions on a graph, you will want precision of 0.01, but if you are simulating data with error, you might round to the nearest integer, in the case where a participant must choose one of the buttons. The response buttons are numbered from 1 to the number of buttons in the response scale, in Recipe_Wiz, so this "rounding" might be part of your theory of how a participant chooses a response.
When the form is filled out, push the "Simulate Many Cases" button to perform the calculations and enter the results in the box. Pushing the "Save" button selects all the data in the window, and a pop-up reminds you to copy and paste these (using CTRL-C and CTRL-V) into some program such as Excel or into a text file, where they could be saved as a comma separated values (.csv) file. The predicted values start in the 12th position and are in the same arrangement as those created by the forms made via Recipe_Wiz.htm. Note that the labels indicate the cells, A1, A2, A3, B1, . . ., AB11, AB12, . . ., ABC111, etc.
In the first column appears the short name of the study, and "date", "time", "IP", etc. are placed in the file where actual dates and times, etc. would have been placed in an empirical study from a form created by Recipe_Wiz.htm. This helps to coordinate the relationship between a file of data collected by a form from Recipe_Wiz, and data simulated by Recipe_Sim. One can then perform analyses on the empirical and simulated data using the same analysis software and settings.
Pushing the "Label Data" will record in the data box the information that the user has used in the simulation, allowing the user to copy and paste this information to keep a record of such information as the names and number of levels of the factors used in the simulation. Because only some of the information is saved, a user doing a number of related simulations might be advised to take screen shots of the window for each setup in order to keep a complete record of the parameters you used in the simulations.