02.01.2013 Views

Physics for Geologists, Second edition

Physics for Geologists, Second edition

Physics for Geologists, Second edition

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Some dangers of mathematical statistics 13 1<br />

a figure <strong>for</strong> the area of 48 641.55 m2. Or the measurements might have been<br />

246.524 and 197.324, and so an area of 48 645.10m2. You may <strong>for</strong>mat<br />

the cell to have no decimal figures, and so believe the figure you see in the<br />

cell, but the computer uses the full quantity unless you tell it not to. At best<br />

you could only say that the area is 48 643 f 2 m2. Not even the last whole<br />

number 3 can be relied upon. You may, however, convert this to 4.864 ha<br />

as the probable area.<br />

The use of prefixes to SI units requires attention in this context. For exam-<br />

ple, 145.53 kN may seem excessive, but if you can measure the pressure to<br />

&4 N it may be acceptable.<br />

In the parameters of mathematical statistics (such as average, standard<br />

deviation, etc.) which by their nature seek to create some order out of dis-<br />

order, the rule of thumb is to use no more than two unless there are good<br />

reasons to use more. But again, your computer will use many more but it will<br />

rarely be worth the bother of doing more than <strong>for</strong>matting <strong>for</strong> two decimal<br />

places.<br />

Linear regression<br />

Mathematical statistics and statistical methods pervade modern science, but<br />

there are great dangers in using techniques that you do not thoroughly<br />

understand. Statistics is a background consideration of the most common<br />

geological operations, such as sample taking, fossil collecting and estimat-<br />

ing the proportions of different constituents in thin-sections of rocks. Is the<br />

sample you have just taken representative? You went <strong>for</strong> the freshest part<br />

of the outcrop - the freshest part that you could reach - so really it is most<br />

unlikely to be representative. But if you don't have a fresh sample, you may<br />

not know sufficiently well what the rock really is or was!<br />

The fundamental problem with statistics in the physical and natural sci-<br />

ences lies in the concept of randomness. When you toss a coin, it is reasonable<br />

to assume that it is a matter of pure chance whether it comes to rest heads<br />

or tails upwards - each is equally likely if the spin of the coin is fast. If large<br />

meteorites hit the Earth from time to time, it would be a reasonable assump-<br />

tion that the rotation of the Earth on its axis and around the Sun, coupled<br />

with the indeterminacy of the positions of the larger meteorites, means that<br />

each square kilometre of the Earth is as likely to receive an impact as any<br />

other. If that is found not to be the case, then one conclusion (there may be<br />

others) will be that meteorites do not approach us from all directions. If they<br />

lie in the plane of the solar system or ecliptic (as they do, f 30°), impacts<br />

will be more likely on a square kilometre in the tropics than at the poles.*<br />

Already the idea of meteorites randomly distributed in space would have to<br />

be qualified.<br />

1 Meteorites have been found in polar snow and ice, so the poles are not immune. Perhaps<br />

some of these would have missed but <strong>for</strong> the <strong>for</strong>ce of gravity.<br />

Copyright 2002 by Richard E. Chapman

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!