time, though, this strategy helps the author, who is very famil-
iar with the data’s complexities, but actually only confuses
readers by creating badly jumbled numbers in the tables.
Always apply the ‘need to know’ criterion rigorously before
accepting any deviation from numerical progression. A numer-
ical progression is desirable in alltables, with only two clear
exceptions: those showing over-time data, and those covering
categorical data which have to be kept in fixed order to be
meaningful (for example, survey response options on a scale
like ‘agree strongly/somewhat agree/somewhat disagree/
strongly disagree’). Other departures from numerical progres-
sion are only very occasionally justified. There might be one or
two cases where readers need to make comparisons across a
small set of easy-to-read tables, and where they would be helped
slightly more by having a standard row sequence across tables,
rather than being given a clear pattern in each table’s data.
In the case of larger tables with multiple columns, achieving
numerical progression is a little trickier. You need to determine
which is the most important column and rearrange the rows so
as to get a numerical progression on that column. Make sure
that the progression column is visible by placing it first (closest
to the row labels) or last (where it will stand out as the salient
column). If you can, try to achieve a numerical progression not
just down the rows but also across columnsin the table, either
ascending (smallest data numbers in the first column and
largest in the last) or descending (largest data numbers in the
first column and smallest in the last). Here you reorder both the
sequence of rows and the sequence of columns to maximize
a table’s readability.^5
Statistics for central level and spread. Table 7.1 provides no
help for readers at all here, but Table 7.2 gives two different
‘averages’, the arithmetic mean, and the median, the observa-
tion coming in the middle or half-way through the data set as
a whole. It also shows the upper and lower quartiles, where the
top quarter and the bottom quarter of the data begin. Readers
can hence see the position of the middle mass, the middle
50 per cent of observations lying on or within the two quartiles.
The data shown here clearly straggle upwards at the top, which
explains why the mean is so much higher than the median,
HANDLING ATTENTION POINTS◆ 169