Thursday, February 26, 2015

GSS 2005 Main File Weight Question


Can WGHT_HSD be only used to create household estimates for variables which are normally weighted by WGHT_PER? (i.e., would they generate inaccurate estimates for variables in sections 10A, 10B, or 11 (the variables for which WGHT_CSP and WGHT_SNT were created?)


Response from Methodology:

The person is considered the basic unit of analysis and many steps have been taken to ensure that the person weight is consistent with external references. It’s the reason why all variables in the data dictionary of General Social Survey 2005 suggest using WGHT_PER, WGHT_CSP or WGHT_SNT for weighting. These are the preferred weights. Sometimes, however, estimates of the number of households will be required, so an approximation of a household weight (WGHT_HSD) has been included on the Main File. It is simply the person weight (WGHT_PER) divided by the number of household members eligible for selection in the survey, i.e. those aged 15 or over. The general recommendation is that GSS household weight be used only for estimate of household characteristics. For example, to estimate the number of households that live in low-rise apartments, WGHT_HSD should be summed over all records with this characteristic.

No household weight was derived for the corresponding person weights WGHT_CSP and WGHT_SNT. As consequence no household estimates can be generated for variables in sections 10A, 10B, or 11.