the difference between co-occurrence and correlation

[eirias]

I have had a nightmarish situation with my data collection over the last week -- I've been able to keep none of the five subjects' worth of data I've collected so far (two T-W, one R-F, two S-M). Three of these were lost for boring reasons that are easily fixed, but the other two got lost because of strange problems with one computer, all of which seemed in some way connected to the display.

• The monitor would turn itself off for 1-3 second periods, for no obvious reason, but it often seemed to happen when DmDX (the experiment software) started.
  
• DmDX kept crashing hard on Session 2's file, and it did this in several different ways -- never the same way twice, the error messages not traceable to anything obvious, but all mentioning something about an object in the display disappearing or dying or something dramatic.
  
• Windows Explorer was routinely failing to refresh, making renaming/moving the data files more hazardous than it ought to be.
  Perhaps importantly, there is a functional clone of this machine on which the experiment has consistently been running fine -- all data lost from that machine has been due to user error.
  
  I talked to Dell and mentioned all these problems at once, saying they might be related and they might not but since a substantially identical machine worked, it sorta sounded like hardware to me. Now, I was using an odd monitor -- one with a touchscreen function, a small difference from the clone machine. I switched monitors and the only problem I could replicate at will, the Explorer issue, still obtained; accordingly the tech support arranged to send me a new video card, which has not yet arrived. Well, it's been a day now and it looks like the most video-card-suspicious issue actually did resolve when I switched monitors. Now what?
  
  It occurred to me later on that another small difference between the clones is when they were set up to run experiments -- could DmDX have updated between those installs? I checked the version numbers and sure enough, the working clone had an older version. I copied the setup files to the new machine, ran the program past the typical points of [insert random display error message here], and it seems to work -- though as with any intermittent problem we won't really know until it has many successes in a row. But I'm guardedly optimistic.
  
  So what about that third error? A little websearch did the trick: some damn cache that's supposed to update itself didn't do it right, which I'm guessing was a side effect of one of the hard crashes the computer took while running the experiment under that username. I deleted the cache, logged out, and logged back in again, and voilà -- Explorer is no longer broken.
  
  The moral of this story, for me: Don't infer that simultaneously-appearing problems actually stem from the same root cause. They may, but it's not a given. (At least Dell makes it easy to return parts you realize you probably don't need!)
  
  Now, here's hoping this next round of data collection is pure success!

Comments

[cognative.livejournal.com]

Ew. That's exactly the kind of computer issues I always hope to avoid. I have a few small ones that are categorized as "mostly harmless" at this point. The most annoying one is that on a Mac my flash drive will fail to work about 1/20 times. It won't work again until I reformat it on my windows machine.

[eirias.livejournal.com]

Wow, that is a high-pressure failure -- must make giving talks a bitch...