The principles here can easily be extended to any kind of
data numbers. Express very large figures in units of hundreds of
millions, or millions or thousands as appropriate. And multiply
very small ratio numbers to get rid of fractions of 1 and the
need for several decimal points. You can also go a long way by
rounding numbers up or down (so that 10.51 becomes 11 for
instance, while 10.49 becomes 10). Or you can just cut numbers
by eliminating all decimal points (which would mean that both
10.51 and 10.49 are expressed as 10). Some people find it help-
ful to design tables using as a rule of thumb that there should
never be more than three ‘effective’ digits in any cell, and
hence no more than 3 numbers vary from one cell to another.
On this rule you might enter 1,215,689 in a table either as
1.22 million, or as 1,220,000 (that is, rounding to the nearest
10,000). If you went to four effective digits the same number
would be 1,216,000 (rounding to the nearest 1000). In any table
showing such large numbers rounding to the nearest 100 is
almost always sensible in cutting away pointless detail, and
often to the nearest 1000. This is especially appropriate where
numbers are being analysed in main text tables, but the same
data are also included in a reference annex or a data CD. Here
there is no need to overburden the main text tables simply in
order to read a precise number into the record.
Numerical progression. The sequence of rows in Table 7.1 is
set alphabetically, so that the data in the second column are
completely jumbled, with one number succeeding another in a
completely unpredictable way. Readers will find the table very
hard to follow, and must fend for themselves in trying to work
out the central level of the data or which health board is doing
well or badly. By contrast Table 7.2 reorders the rows to give
a clear downward numerical progression. Health boards’ per-
formances here are visible at a glance, with strongly performing
boards at the top of the table and weakly performing ones at
the bottom.
Never keep data arranged in alphabetical ordering of rows or
some other customary order if this obscures the numerical
progression in the table. Some authors argue against this advice
because they want to present data for cases or other units in the
same standard sequence from one table to another. Most of the
168 ◆AUTHORING A PHD