... personal wiki, blog and notes
Bryan's Blog 2009/07/16
Bryan's Fishy Example
I've had a few days lately where I've trotted out my favourite "easy to understand example" as to why some scientific datasets need to be preserved, and some do not. In the final analysis, we find that it's easy to identify some datasets as "must preserve" and then we enter "value judgement" territory. But it's helpful to have a strawman to consider on the way.
With apologies to those of an aquatic bent, I like to pick on fish.
Imagine that I make a series of complex measurements of the anatomy of my favourite fish using my equally favourite instruments. I publish those measurements. Should those data be actively managed for posterity by some third parties (other than me)?
Well, I would answer no, with a couple of caveats we'll get to. For a start, someone else could wheel out their favourite instruments and make the same measurements on another example of the species1. This of course assumes a) there are enough of them around that more can be found, and b) they're easily accessible (eg. not at the bottom of the Mariana trench). If either of those assumptions are broken, then maybe my measurements might be nearly as important as I think they are (like all scientists my natural assumption is that any measurement I make must be really important). We'd better also add the caveat that if we can predict the usefulness of these data in some composite study, then we might want to have it managed too ...
On the other hand, if I made measurements of the number of this fish in some geographical region, that measurement can be repeated for sure, but it will be at a different time, and the time sequence will in and of itself be another interesting measurement. So by default, in this case my (original) data will be (or could be, there are few guarantees in this game) of use in some future study.
Clearly the difference between these two cases (sans caveats) is that in first instance, I was making a measurement that was repeatable (and the likelihood is that the repetition would be in the future with better, stronger, faster instruments). There is no intrinsic interest in joining those two sets of data (my original and the later better) to build a newer more useful dataset. Nor, I presume, is it likely that my numbers will form an input into some later calculations. In the second instance, it's obvious that my data could form part of a useful composite dataset, and it's obvious from the moment I made the measurements.
So, now let's abstract that a little bit. We have two sets of data:
typeB: immediately and obviously of potential interest in the future, and
typeA: not obviously a candidate for future composite analysis.
Given we (humanity) have a finite resources for data management, it's a no brainer: we manage typeB, and consider typeA for a bit longer.
typeB was easy, but for typeA we have a few things to think about. We've got our caveats:
If it's likely because of rarity or difficulty in obtaining samples, that the typeA measurement can't be repeated: Manage it. If not, ask the next set of questions:
Is it likely that some sort of composite dataset will be constructed using my data (perhaps we can foresee the necessity to build the statistics of these anatomical measurements across the population)? The keyword here is likely and it has to be weighed up against the difficulty, expense, and likelihood of repeating the measurement. So here we are in value judgement territory.
Some of these value judgements are obvious, some are less so, but at the end of the day it's pretty hard to set up a hard and fast set of rules. So it's a continuum: at one end, we just make a rule, manage that type of data (e.g. gene sequences), at the other we decide not to bother (simulation sensitivity analyses). In between, we dither. For the "not to bother" stuff we shouldn't do nothing though: that material is suitable for personal data management.
What? Personal data management? Has Bryan been snorting something naughty? (Editor: no).
None of the above considers the issue of "preserving the scientific record". Regular readers will know that I consider that a pretty important issue, and one that also leads to an argument about keeping data. However, I think much of the typeA data above is more like the lab book you (should) keep: it should exist in your possession for some period (years) after you publish the data, but at the end of the day, it doesn't really need to persist. In this case, the scientific record will get on just fine without the numbers. TypeB data, though, we need to keep it, not for my sake, but for all our sakes.
Before we leave this example though, for completeness, we should consider one more issue. If, when I made my anatomical measurements, I tagged them with the geographical location from where I nicked my species examplar, then I'm vastly increasing the chances of my data being useful in some environmental application. So the lesson there is, the more complete my metadata is, the more likely my data is to become typeB and an obvious candidate for future use. Now, when we get data citation working, it also becomes a candidate for citation independent of the paper I wrote about my fish. Collect and write that metadata. You will get the credit one day!
Reading in 2009, 14: Canadian Wildish West
I can see why this was a prize winning book. It is a good read, although I thought a few too many threads were left hanging at the end. In that sense it felt not quite finished, although to be fair, the hanging threads weren't dangling completely unsuspended. Read it for yourself to make sense of this paragraph!
It is worth the read. The book is set in Canada in 1867, where in a small community a man is found dead, scalped, in a cabin. Much follows, with several treks across the wintry desolate Canadian north (probably by a few too many people, there's one lot whose only function seems to be to link two disparate parts of the story). There are some characters who are drawn in lots of detail, and then seem to disappear without trace, and others are drawn in less detail, but more central to the story. However, all this seems a bit carping. I liked the book, and as a tale of epic derring do of (wo)man against nature, it does very well.
I'm amused that wikipedia says that Penny was suffering from agoraphobia while writing the book, and the research was all done in the library. What was that about plagiarism being copying from one person, and research (novelty in this case) being copying from many people?