2003 Scottish Health Survey Health Board Tables - Survey Sample Size and Precision
As surveys are carried out on a random sample of the population, there is always some uncertainty about the survey estimates. Everything else being equal, the precision of an estimate increases as the sample size increases. When interpreting a survey estimate, therefore, it is important to also consider the uncertainty of that estimate, represented by the standard error (or confidence interval). This becomes particularly important for estimates based on sub-groups of the population, as the sample sizes can become relatively small.
As well as being related to the sample size, the precision of survey estimates is also affected by the design of the sampling. Clustering of the sample and unequal probability selection (and weighting due to non-response) both impact on the precision of the estimates - usually to reduce the precision. In addition, the stratification employed when obtaining the sample of primary sampling units also impacts on the precision - usually to increase it. One method of reporting the impact of these design features on the precision is to report design factors (defts); the design factor is calculated as the ratio of the 'true' standard error for the survey to the standard error that would be obtained with a simple random sample ( i.e. no clustering, stratification or weighting) of the same size. The larger the value of the deft, the greater the loss in precision, with a deft of 1 indicating no loss in precision.
Given that the estimates by Health Board and gender can be based on small sample sizes, we have generated estimates of the standard errors, defts and 95% confidence intervals for a selection of survey estimates (means and proportions), so that, in combination with the sample size, the true precision for each estimate can be assessed. Note that the deft values and true standard errors are themselves estimates subject to random sampling error.
Volume 4, Chapter 1 of the 2003 Scottish Health Survey report has more details of the sample design and weighting.
If you have any further queries about these Health Board tables, or the Scottish Health Survey, please email: shs@scotcen.org.uk
List of Tables
Any CVD, IHD, IHD or stroke, and Diabetes, by Health Board and sex
Estimated usual weekly alcohol consumption level, by Health Board and sex
Units consumed on heaviest drinking day in past week, by Health Board and sex
Cigarette smoking status, by Health Board and sex
Fruit and vegetable consumption (portions per day) by Health Board and sex
Body mass index ( BMI), by Health Board and sex
Waist-hip ratio ( WHR) and waist circumference, by Health Board and sex
Self-assessed general health and prevalence of long-standing illness, by Health Board and sex
GHQ12scores, by Health Board and sex
Dental health, by Health Board and sex
Annual accident rate, by Health Board and sex
Wheezing in past 12 months and doctor-diagnosed asthma prevalence, by Health Board and sex
FEV1, by Health Board and sex
Blood pressure level (140/90 mmHg definition), by Health Board and sex