Random Distribution

I'm using the RWG parameter with unconditional skips in order to randomly assign subjects to conditions (i.e. to expose them to different stimuli). Since there are only two conditions, ideally I would have close to an even split between conditions. However, I noticed that I have more like a 3:2 split after running the experiment 60 times.

I understand that random means random and, though unlikely, I could potentialy have a 60:0 split. Just wondering if there is anyway to assure myself of close to an 1:1 ratio.

Thanks,
Jason

In some cases, unfortunately, random does mean random. You can get around this by coming up with subject and condition designations ahead of time (by whatever method you like) and then creating an [I]autostart.txt[/I] file.

See if this helps: [URL]http://www.empirisoft.com/medialab/help/preferences.htm#prefs_auto[/URL] (scroll to the end of the section when you get there)

Jarvis,

Thanks for the response. I have already started to use another assignment method.

The autostart option would have been great for my method if I were using just one computer. However, since I am running multiple subjects at one time I would need an implementation that marked the autostart file as soon as the experiment is begun. The way it works now, all subjects in the same session would have the same id and same condition.

I hope that Empirisoft eventually develops a network friendly version that would support multiple subjects running at once. For now, I can get by with my tricks to make it work.

Care to share your tricks? Perhaps they could inspire a general purpose network solution. That, and just curious :)