HANDLING ATTENTION POINTS◆ 191Median smoothing twice will get the change back to taking
place at its real time in May.)
To see how median- and mean-smoothing work look at
Figure 7.5. The chart shows some opinion poll figures I have
made up, purporting to show the proportion of UK citizens
who believed that Tony Blair should become President of
Europe in 2001, with median-smoothing applied. The
smoothed series is shown as the solid line here, with the actual
data observations as a thinner dashed line, a technique which
allows readers to focus most on the smoothed trend but still
retain the ability to see how the actual scores moved over time.
Observations for very unusual months show up very promi-
nently as big divergences between the two lines, inviting you to
give a special explanation of them. (One small digression point
on methods here. You will need to have data for a few observa-
tions before and after the period you want to look at, in order
to be able to get smoothed data covering the whole period you
are interested in. There are techniques for finding starting and
finishing values for smoothed series where you do not have this
extra data.^8 )
010203040506070JanuaryFebruaryMarchApril MayJune JulyAugust
SeptemberOctoberNovemberDecember
Month% saying Blair should be Euro-PresidentRaw score Smoothed scoreFigure 7.5 An example of median-smoothing – the
percentage of British voters saying Tony Blair should
become President of Europe during 2001 (hypothetical)