Herefordshire Young Peoples Survey 2000
Synopsis:
Although we sampled nearly every Herefordshire school, these schools
are in a statistical sense only a sample from a much larger population
of Year 7, 9, or 11 children in Herefordshire. Because they are a
sample, by chance alone, that sample could have a slightly different
prevalence of x to the true population prevalence, the prevalence in the
much larger population. Just because we see a difference between 1996
and 2000 in the prevalence of x, does not mean that the true prevalence,
the prevalence in the much larger population, has really changed. We
have used 95% confidence intervals (95%CI) to help show this. Our tables
show the prevalence of x in 1996 and in 2000 and the difference between
these with the 95%CIs of the difference. If the 95%CIs do not include
zero, we can conclude that chance (variation because we sampled from a
large population) is an unlikely cause of the difference and the
prevalence of x might really have changed. If the 95%CIs do not include
zero, we must conclude that chance is a potential cause of the
difference seen and the true prevalence might not have changed.
However, the situation is more complicated than this. We have
measured the prevalence of over 200 attitudes and behaviours. Even if
the true prevalence of all of them were exactly the same in
Herefordshire, it is almost certain that some prevalences would have
95%CIs that do not include zero and appear to have “really changed”.
This is a complicated point. However, it is akin to the situation where
you buy every ticket in a raffle. It is unlikely that any specific
single ticket (e.g. number 27) in your hand will win, but certain that
at least one of them will. By testing lots of differences, it is likely
that by chance alone, you will find an unusual finding, which is all
that the confidence intervals are trying to tell us. So, if you find
that the prevalence of x has changed, look for related items y and z. If
they have moved in the same direction, then you can be surer about the
true change in x. If y and z have not changed, it could be that the
“statistically significant” true change in x is simply an unusual
occurrence, and not a true change at all.
To download the full report please
click
here
For further information please see:
www.herefordshire.gov.uk |
