CLUSTER is a 5-digit numeric variable reporting the scrambled pseudo primary sampling unit (PSU) cluster variable.
99999 = Not generated
CLUSTER is the scrambled pseudo primary sampling unit (PSU) cluster variable designed for use with STRATA in Taylor series linear approximation for correction of complex sample design characteristics. See the STRATA variable description for more details. These variables are mainly used to group observations into PSU clusters when computing standard errors.
Sampling weights will often produce correct point estimates (such as means and variances). Some researchers may use additional techniques for the purpose of analysis and variance estimation. ATUS has provided STRATA and CLUSTER variables for this purpose.
In the Stata software, researchers may submit the svyset and svydes commands once to define these variables in a data set (including sampling weights such as WT06). Other software may require that CLUSTER and STRATA be identified for each individual analysis.
Users should note that STRATA and CLUSTER are not available for ATUS respondents whose final CPS interview was in April 2014 or later; these cases are assigned a value of 9999. To generate correct standard errors from April 2014 forward, users should use replicate weights (RWT06).
- All households.