`HANDLING ATTENTION POINTS◆ 181`

names on the horizontal axis, and so a separate key is needed

to show which number denotes which board. Where bar labels

are lengthy, always choose a horizontal bar design, like that

shown in Figure 7.3 where the health board names are easily

accommodated. Aim to use a fully informative label wherever

possible, with minimum abbreviation. This approach follows

the one-stop look-up principle discussed earlier in connection

with referencing systems.

Numerical progression. In Figure 7.2 the bar chart format cuts

out the mass of details in Table 7.1. But even so without any

pattern across the bars, the chart in Figure 7.2 is a jumble of data.

By contrast Figure 7.3 reorders the bars in a descending sequence,

showing completely clear results. The median and the two quar-

tile bars are also indicated, which would not be feasible without

a numerical progression. In all charts (except those showing

over-time patterns or categories where the sequence of values is

fixed) achieving a numerical progression is just as vital as for

tables.

Showing specific numbers. In Figure 7.2 the choice of a narrow

vertical bar layout and the use of an index of cataract operations

per 1000 population with very large data numbers makes it

impossible to show any numbers for the bars. By using a hori-

zontal bar layout, and an index showing cataract operations per

100,000 people, which generates simpler numbers, Figure 7.3

can give precise numbers for all observations. Note that these

numbers are included withinthe bar space. Avoid adding num-

bers above the bar area with vertical bar charts, or to the right of

the bars in horizontal bar charts, because in both these cases the

number will detract from the proper visual scale of the bars.

Although Figure 7.3 has an appropriate number of gridlines and

tick points for readers to be able to scale the bars, including the

numbers removes any difficulty in readers having to estimate

what the individual scores are.

Scaling and grid linesdecisions are often messed up. The two

figures here are both scaled fairly well, but the vertical scale in

Figure 7.2 could have been greater to allow more variation

amongst the small scores to be seen. With more extreme ranges

in the variation of data it is common to see charts where

the vertical scale for the bars has been set automatically by the

spreadsheet. This may highlight unusually high or low observa-

tions, but at the price of making almost invisible patterns in the