Running an Experiment: Technical
	chil home > lab only > polocies and procedures >

On the server

Consider this a checklist, and please be sure to keep everything updated.

Experiment Number

Before you do anything else, you need to get a lab experiment number. (This is distinct from a departmental or Experimetrix number.) Go to the file "Red Giant:X Support:Experiment Info.xls" and grab the next available number. Be sure to fill out the other relevant information, too.

Experiment folder

Each experiment should have its own folder on Red Giant, inside the "X Support" folder. Inside that folder, there should be at least four things:

• A folder containing all the relevant paperwork, such as the consent form, the feedback forms, and any forms needed by the experimenters.
• A folder for all the data files.
• The experiment code, named something obviously tied to the experiment number, along the lines of "x25.lisp" or some such.
• A folder containing all development work, including additional source code and any and all subject instructions, picture files, and the like.

It is crucial that all this information be maintained on the server so that experiments can be re-analyzed and replicated later.

Questionnaire

Each experiment also needs a questionnaire. These are kept on the "WebServer" network volume on the server, in the folder "Documents:OpenForm:" Each experiment should get its own folder. There are two ways this can be handled, described below.

The Current Way

The easy way is to use the most recent experiment code, which automatically creates records in the database with experiment number, subject number, and condition number. This version also supports different experiment-specific questions for each condition.

Under this scheme, you need one HTML file per condition in your experiment for the experiment-specific/condition-specific questions you want asked. If you have only one condition, then the file should be "c.html"; if you have multiple conditions, you want "c.html" for condition 0, "c1.html" for condition 1, "c2.html" for condition 2, and so on. Note that these files can be exact duplicates of one another if you have the same questions for subjects in all conditions.

You can find a basic template for the "c.html" in the folder "xTemplates". You should not need an "a.html" or "b.html", etc. folder if you're doing things this way.

The Old Way

Don't use this unless you have a strong reason do to so. In this scheme, each experiment folder should have four files in it, "a.html" through "d.html". Templates can be found in the "xTemplates:retired" folder in that directory. Note that you must change the experiment number in each of the four HTML files to match your experiment number or things will get messed up. (The spot you need to change is noted as "****SETTHIS****" in each file. Also note that as currently configured, the server can process up to 10 user variables to record information pertinent to your particular experiment, though this can be changed to accommodate more. Talk to Mike if you need to do so.

Setting up to run

Make sure all the relevant files are copied to the local hard disk, especially the experiment application. Every experiment should have its own folder on the subject running machines, and it should be there. DO NOT RUN THE EXPERIMENT APPLICATION OVER THE NETWORK. This will cause it to run poorly and the timing to be off.

Experiment applications will (usually) be written in MCL. To run MCL, the application needs to find certain files. These files are:

MCL-compiler-5.1
MCL-kernel-5.1
MCL-library-5.1
pmcl-OSX-kernel

The application will look for these in the same folder as the MCL application itself, and in the "Extensions" folder inside the active System Folder. These files should be in the MCL already. If you get an error message saying one of these files cannot be found, please copy all three of these files from Red Giant to the machine's System Folder:Extensions folder.

(These files live in "RedGiant:Public:Installers:MCL 5.1ƒ:MCL-compiler-5.1" etc.)

The other thing you need to do is look at the "bincarbon" folder inside the MCL folder on the eMac. This should exactly match the "bincarbon" folder on the server in the MCL 5.1 folder there. If not, copy files over from the sever until the two folders match up.

Before you start running real subjects, please test both the experiment and the Web form thoroughly. Lost data because of sloppiness is really bad.

Running

Once the application has been launched, it will pop up a window entitled "Listener." You invoke the experiment by first loading it. Go to "Load File..." in teh "File" menu and select the relevant file (probably "xNN.lisp" where NN is your experiment nubmer), which you should have copied to the local disk.

You run your experiment by typing (begin-experiment) and then hitting <enter> or <command-e>. You will be prompted for various things, like the subject number, where data should be saved, and the location of the instructions file. Once the instructions are displayed you're ready to go.

If any problem occurs during the experiment and Lisp generates an error that is displayed on the Listener, save the Listener window to a file and email that file to whoever programmed the experiment. It is very important that you do this so the problem can be tracked down.

You may be able to continue the experiment, but then you may not--at least try to do so. If it's early in the experiment and you think you have time, you can try to restart from the beginning, but don't be too aggressive about this.

Once your sessions for the day are finished, all data should be manually backed up from the local machine to Red Giant in the appropriate data folder for the experiment.

	People   Facilities   Research   Projects   Miscellany   Search   |   Home

Last modified 2007.06.08