Methods Used by SDA 3.5 for Computing

Standard Errors for Complex Samples


CONTENTS


Common to All Analysis Programs

Specifying the Sample Design

The sample design is specified for each study when the SDA dataset is defined in the Hypertext Archive Definition file (HARC file). In that file, the owner of the data archive (or whoever can modify that file) may specify a stratum variable and/or a cluster variable for each dataset.

Consequently, there are three possible specifications:

Each of these three specifications results in standard errors being computed differently.


Domains and subclasses of the sample

When analyzing data, we often need to calculate statistics for subgroups of the sample, in addition to calculating statistics for the sample as a whole.


Methods Specific to Certain Programs


Differences between SDA and other programs


References

For a more detailed technical treatment of the standard error specifications for the SDA programs, see the following documents: