Methodology

Survey design, sampling methodology, and statistical analysis approaches

Survey Design

Children's Dental Health Survey 2013

The 2013 CDHS used a multi-stage stratified cluster sampling design:

  1. Stage 1: Primary Sampling Units (PSUs) - regions and local authorities selected with probability proportional to size
  2. Stage 2: School groups selected randomly without replacement
  3. Stage 3: Schools within groups by simple random sampling
  4. Stage 4: Children within schools by sequential random sampling

Adult Dental Health Survey 2009

The 2009 ADHS employed a stratified two-stage cluster sampling design with postcode sectors as primary sampling units and addresses as secondary units.

Survey Weights

Survey weights adjust for unequal selection probabilities and non-response. Different weights are provided for different types of analysis:

Weight VariableUse For
eweightClinical examination variables (e.g., DMFT, decay counts)
ppweightPupil self-completion questionnaire variables
prweightParent/guardian questionnaire variables

Statistical Methods

Survey-Weighted Analysis

All analyses in XeraCDHS use survey-weighted methods that account for:

  • Complex sampling design (stratification and clustering)
  • Unequal selection probabilities
  • Non-response adjustments

Available Regression Models

Linear Regression
Continuous outcomes
svyglm(y ~ x, design, family = gaussian())
Logistic Regression
Binary outcomes (0/1)
svyglm(y ~ x, design, family = quasibinomial())
Poisson Regression
Count outcomes
svyglm(y ~ x, design, family = poisson())
Negative Binomial
Overdispersed count outcomes
svyglm(y ~ x, design, family = quasipoisson())

References & Documentation