Abstract:
|
Analyzing surveys utilizing complex sample designs is different from that through simple random sampling. When performing the complex survey data analysis, the final calibrated weights and the complex sample design need to be taken into consideration. Information about the complex sample design may sometimes not be considered by the analyst, especially when the sample design is not clearly documented or the use of replicate weights is not well understood. This could lead to incorrect variance estimation and a loss in accuracy of the results. To prevent this, a combination of "replicate weights" & "calibration weights" during the complex survey analysis is recommended. Many national surveys provide a set of replicate weights to estimate the variance while accounting for the sampling design. However, not all researchers may be aware of the value of incorporating replicate weights in their smaller scale research projects or in their analyses using commercial statistical software like SPSS. In this presentation, a brief introduction of replicate weights will be given followed by a demonstration on how to perform different types of analyses with replicate weights using the SPSS macro.
|