Vaguely Hopeful, Curve-Fitting Edition

One of the projects at work might actually be yielding something potentially interesting. So all is not lost (yet).

When there's nothing to clean at work, nobody ever thinks of any work for me to do, so I spend the time curve-fitting the painstakingly gathered data (it's either that or twiddle my thumbs - no surfing the Internet on company time!). Since the samples are wildly variable in their properties, the curves aren't expected to fit very well*. In fact, my coworkers probably think that what I do is a waste of time. (Now if they actually *have* something for me to do, I won't be "wasting" my time...).

Anyway, I think I've actually got something that's surprisingly nice. R² > 0.9, for the most part, and visually the curves seem to fit, also. I realize that R² values can be horribly misleading, and I might've calculated it wrong, but I don't know a better way of assessing fit because of my aversion to statistics. (Too bad I am not dealing with probability distributions.) Don't know *why* the data follows such a curve, though. Something to show my coworkers on Monday, perhaps.

Reminder to self: try to automate the data analysis process. (Ick, programming...but there's no denying that things will go faster if the code works.)

Random background:

The project** basically consists of subjecting the samples of interest to strong external forces, and see what happens. It belonged to a previous student and I am supposed to continue the work. The question is, what work, precisely? My coworkers are all pretty "experimental" people (*mutters about chemists who work near mechanical engineers for too long*); previous work on the topic consists of slowly gathering data and plotting them randomly in hopes for some correlation between different factors, then bullsh- making up elucidating the reasons behind the correlations. It's such a haphazard approach, and I don't like it. (Also, some of the observed phenomena are actually fairly well known, and well, why re-invent the wheel?)

Anyway, if one has discovered the correlations between the various parameters and explained them, what more is there to do? "Collect more data", they say, but to what end? What should be done with the data? (Yes, curve-fitting is my answer to this question, but it's a pretty limited answer.)

Some new paths include applying the force in slightly different ways (e.g. constant, piecewise, or oscillatory instead of logarithmic or linear) and see what happens. (The answer to this question, so far, is "nothing much", by the way. But we'll see.)

*At this point, some people will probably say: "welcome to real life". To this I say: real life sometimes matches with theory gorgeously well, too! (E.g. many calibration curves are beautifully linear. *Sigh*, I miss analytical chemistry.)
**I really should think up some handles*** for the projects since there's more than one and now everything is referred to as "the project". Hm.
***Too much Matlab. Bah.