I am looking for some advise on how to analyse data from several repeat
surveys that collected the data using a quota sampling approach.
To get a concept of precision within a probability sample you would
obviously get the confidence intervals based on sampling errors taking into
consideration stratification or Primary sampling units if this was not a
simple random sample. The methodology would extend to testing differences
between surveys/trend tests. However, theoretically or practically how
would you undertake such analyses if the data you had was collected via
Could you use some bootstrapping technique to produce some type of
confidence intervals. If so how would you extend this to testing
differences between two surveys.
I would be interesting to know what happens in practice. I note that
opinion poll results have levels of confidence attached to some estimates
but I can't remember any reported statistical analyses of these that looked
at changes across time, although I may have just not be interested at the
All thoughts greatly received and if there sufficient they will be
summarised and posted back to the list. However, responses will be treated
in strictest confidence if required.
Looking forward to your replies,
Dr Simon Clarke
Epidemiology and Medical Statistics Unit
Health and Safety Executive
Room 243A Magdalen House
Merseyside L20 3QZ
Tel: +44 (0)151 9513832
Fax: +44 (0)151 9514703
E-mail: [log in to unmask]
Most of our statistics are available online
The original of this email was scanned for viruses by the Government Secure Intranet (GSi) virus scanning service supplied exclusively by Energis in partnership with MessageLabs.
On leaving the GSi this email was certified virus-free