01 December 2005

Data from the backyard

Measuring snail shells, not just one or two, but lots and lots of them, is a passion of mine. Measuring as many specimens as possible (up to a limit) provides statistically more meaningful data. But I don’t like killing large numbers of snails and refrain from doing so if I can. One good argument against permanently removing large numbers of live snails (or any other animal) from a colony that is under long-term surveillance is that the removal of animals could unnaturally influence the future values of the characteristics that are being studied. So I usually measure empty shells and wherever there are large numbers of empty shells of one species, I try to collect as many as I can carry.

I have written about the tiny snail Vertigo pygmaea that lives in my backyard. Actually, they live not just in the backyard, but all around the house. Having them near my house is convenient, because I can collect even live snails, measure them and then return them within a day or two to where I got them from. That way, the snails are not harmed and I have my data. The graph below shows some of the data I have obtained with V. pygmaea since 2001.

These are "box and whisker" plots for shell heights of Vertigo pygmaea from my yard measured on 3 occasions. The top and bottom of each box are the 75% and 25% quartiles and the box encloses the middle 50% of the data for a sample. The horizontal line inside a box shows the median. The horizontal lines (“whiskers”) end at the minimum and the maximum values (there are other ways of determining where the whiskers will end). The numbers above the boxes are the number of specimens measured in each sample. I drew this plot using PAST, a free and user friendly statistics program specifically designed for paleontologists.

This graph shows that the median shell length has been stable during the study period. Although, because of the enlarged span of the horizontal axis, the median values seem to vary, the actual values, 1.89, 1.90 and 1.87 mm, can all be rounded off as 1.9 mm.

Continued here.


pascal said...

Thanks for the link to PAST. What's your technique for measuring shells? And, what's your upper limit to sample size?


I measure the smallest shells under a dissecting scope with an eyepiece reticle. I have experimented with another method that involves photographing the shells first and then measuring them later on the computer screen. You also have to measure some sort of ruler with the same set up. But I haven't implemented the 2nd method yet. If I ever do, I'll write about it here.

There is no one definite upper limit to sample size (papers have been written about it). It all depends on why you are measuring the shells. If you only want to get the mean & the standard deviation, about 40 or so shells should be good enough. I have done some simple simulations. I can't remember the exact numbers, but after about 40 or so, the mean & SD get very close to population values. It'd be a total waste of time to measure 200 shells just to get the mean & SD.

But if you want to see the shape of the distribution, then you need much larger samples: 100, 150, 200... But you probably wouldn't need to measure 1000 shells.

Christina said...

I'm a US student working on a project in Botswana now on snail habitat distribution. I would like to include snail sizes in my data. Can you please let me know how you measured the sizes of your snails? Not the technique...but from-where-to-where? thanks


Thanks for the question, Christina. I have posted a reply here.

Good luck with your work. it sounds interesting.