Lab Group Assigning

• A tool for teachers to assign students to lab groups written in Processing's java code Processing. It may be built for running on Apple OSX, Linux and Windows operating systems.
• This tool selects groups of student teams from a student class population to be involved in lab tasks during experiment rounds. Each student is involved once every session round and once for each different tasks.
• Each lab group, i.e. a combination of students as a team, like a student pair for example, is assigned no more than once. The number of possible lab groups, referred to as Pool Size, is the number of group sized student combinations chosen from the class size.
• The output results go to both the console and the Processing application screen.
• The inputs are:
  • Class Size - The number of students in the selection pool.
  • Group Qty - The number of tasks. These are the output matrix columns.
  • Rounds Qty - The number of session rounds. These are the output matrix rows.
  • Group Size - The student grouping size. The matrix cells are the assigned student groups. The students are selected from the class, i.e. the selection pool. The integer numbers identify the students. For example the student group, "03,06,12", is a group consisting of student 03, student 06 and student 12.
  • Auto Save Solutions - A checkbox for running a series of solutions for the same conditions where each solution result is automatically saved to a text file.
  • Auto Saved Solutions - The number of solutions that will be run with the same conditions where each solution result is saved to text file.
  • Do A Proportion Estimate - Option to perform a 95% confidence interval for the proportion of assignment solutions there are in all the possible assignment matrix permutations.
  • Solution Qty. Basis For Estimate - The number of solutions to find in order to make the solution proportion estimate. The minimum value is 5.
  • File Est. Prop. Progress - Creates a tsv text file listing the estimated proportion information associated with each found solution in the basis for proportion estimate. This information can be graphed to observe the proportion estimate history for ever increasing solution samples found.
  • Run Ruthless - Option to run faster.
• The Class Size should be equal to the Group Size x (larger of Group Qty and Rounds Qty). Student selection will not be a balanced participation when Group Qty and Rounds Qty are not equal. The same applies when the Class Size is larger than the required number. Chk value variation indicates unbalanced participation.

Example Screen - Solution Found

Example Screen Output - Keypress Terminated Run

Example Screen Output - Proportion Estimate

• The displayed lab groups matrix and the trial history data always corresponds to the last trial run. In the above image for a proportion estimate session that was to find 100 solution assignments, the last solution found required 941 trial guesses taking 0.112 seconds. The entire session of 100 found solutions required 1,151,365 trial guesses (samples) taking 1.7275 minutes.

Notes

• The application window is re-sizable at any time. It also starts up running a default assignment that may be still running by the time you want to interact with the program. Press the q key to halt the current process. On fast running computers the button controls require a longer press to activate than what you might expect.
• Control options indicated in Red color are not available for use while a process is running or are not applicable in context with how other selected control options.
• Execution time is pure luck for the lengthy tasks.
• The Chk: values are row and column checksum values that are the sum of the assigned student id numbers. Properly assigned row and columns will have the same checksum. The Chk: values play no role in the process. They were originally provided during development for quickly spotting solution errors, then subsequently found to be a useful visual confirmation.
• This program continuously makes random lab group selections for each matrix value from an ever decreasing pool of remaining possible lab group candidates. It records the most filled assignment matrix and tracks the least unfilled number score. Matrix build attempts are abandoned when it develops an unfilled score equal to the best and therefore cannot be a displayed improvement. The displayed matrix is the last best assignment matrix build. Selection trials stop when the matrix is totally filled or when the attempt number reaches a maximum limit. That limit is arbitrarily set to 1 trillion attempts.
• Matrix values like --,--,-- indicate there was no valid lab group combination possible for that matrix position for the trial run.
• Option Run Ruthless sets a faster running mode. It cancels a trial at the first impossible matrix group combination position instance. Without Run Ruthless the process would continue to guess a failed matrix that scores better than the current best scoring failed matrix. Displaying subsequent "better" matrices has an unusual effect. It makes the process look like there is actual progress going on.
• Terminating a long running solution attempt and then restarting it gains nothing. It would be pure random luck that a restarted solution attempt can quickly arrive at a solution. The matrix displayed while running a solution attempt is only the first best solution so far. It is not the basis for any subsequent possible solution. It represents a random matrix guess that failed because it was impossible to fill the unfilled cells. For every unfilled cell you can check and see for yourself that there could not be a remaining group that could satisfy the selection criteria.
• The time required to complete the assignment matrix is a function of the selection pool size. The selection pool size depends on the class size (n) and the group size (k) as n!/k!*(n-k)! . The number of possible assignment matrices of size gr (group qty x rounds qty) that can be selected from the pool size p is p!/gr!*(p-gr)! . That number can be enormous.
• The number of valid solution matrices within the pool of possible assignment matrices is what the proportion estimate is about. The valid solution number is often large enough to quickly guess a valid solution at random. For a class size of 10 students with 2 students per lab group there is a pool of about 3.17 trillion possible ways to fill the 5x5 matrix with lab groups. The proportion estimate for solutions in that pool of 3.17 trillion is around 0.003676 or about 11.65 billion solutions.
• Pressing the q key stops the process and outputs the current best solution. The Stop button does the same but it may not be as responsive as the q key press.

File Button

• The File button creates a text file of the current results within the folder that contains the application. The files saved are named "LGM_<Class Size>_<Group Size>_<Group Qty>x<Rounds Qty>" with a time code indicating the time the file is created. For example the file named LGM_10_2_5x5_2021-6-1_15447_269223+88.txt was made at 15:44:07 and some milliseconds + a random number on June 1, 2021. The millisecond + random number provides for unique filenames when files happen to be made very quickly using the Auto Save Solutions option when the solution conditions solve very quickly.
• Use the Auto Save Solutions checkbox and the Auto Saved Solutions quantity entry field to automatically save results.
• When the Report Unused in File option is checked the File output reports the student group combinations that were not selected for the lab groups matrix. On Linux systems, where one runs the application from a Terminal window, the unused groups list outputs to the Terminal window whenever a trials session completes or one stops a running process.
• The File button is enabled and functional at all times. When pressed during a process the resulting file is a snapshot of the current state.

Install and Running

• A Prebuilt application file for Apple OSX is provided as a zip file is available in the releases section.
• Prebuilt application files for Linux and Windows are not provided in this repository. LabGroupAssigning is easily built from the Processing IDE (version 3 or 4) using the files in this repository. LabGroupAssigning can be executed from within the Processing IDE. Those familiar with Processing would need no instructions other than the knowledge that the Peter Lager G4P GUI controls library needs to be installed. Those not familiar with Processing need to start here: Processing where you can easily download and install it.
• Processing projects consist of .pde text files within a folder named to match that of the main project file. If you are downloading the LabGroupAssigning repository, not the prebuilt application, there will be a folder LabGroupAssigning inside the repository folder LabGroupAssigning. That inner LabGroupAssigning folder is the Processing project that contains the .pde files. Open any one of those .pde files from within the Processing IDE will load the LabGroupAssigning project.
• A Processing application is a Java application. Therefore the Java run-time system is required to be installed or packaged within the application. The option to package Java within the application is possible only with Apple OSX systems. Java will need to be installed on Linux and Windows systems.
• Window Un-Minimizing (Unhide) on Apple OSX: Select the running application's icon in the Dock to restore the minimized window. The exported Processing application for Apple OSX has a simple top menu that lacks Window controls used to restore a minimized window one might happen to minimize via the application's Yellow control dot. Command Tab switching to the application will not resize the minimized application window. This is normal behavior for OSX applications, but OSX applications have the top menu Window control where the OS places a restore option for all minimized child windows. As of this writing exported Processing 3 applications lack that control.

Console Output

• This application continuously writes to the console as it runs. The console output is similar to GUI output but captures more information in certain circumstances because the GUI writes over itself faster than what you can notice. The console scrolls instead. The proportion estimate improvement over sample trials is observable in the console. Console output is shown when running the application from a Terminal session. In Linux you run the application from a Terminal session, so you get to see it automatically. Starting the application from a Terminal session in Mac OSX is accomplished by running the LabGroupsAssigning file within a Terminal session. This file is at the folder Contents/MacOS within the MacOSX LabGroupAssigning application package.

File Est. Prop. Progress

• The following images are graphs created from a tsv text file listing the estimated proportion information associated with each found solution in the basis for proportion estimate.