Why dont the CSPLAN commands specify without-replacement (WOR) sampling?
It is possible to add this term, but it has little practical effect on standard errors. In our trials with Statas svy suite of commands, standard errors deviated at only the 3rd decimal place when specifying a finite population correction for the MQ dataset (computed as the number of achieved observations divided by the population total, with the latter indicated in Table 2.1 in the WERS 2004 Technical Report). Use of the term is also questionable when using non-random subsets of the full data set. It might be noted that NatCen (the fieldwork company for WERS) did not use this feature when computing design effects for the WERS 2004 Technical Report. How does SPSS estimate variances in the Complex Samples Module: This is not specified in the manual. However, the speed of estimation suggests that variances are estimated via linearization rather than through replication methods such as the jacknife. Sample output: The sample output provided in the table below includes a worked example of
Related Questions
- I seem to remember that the facility to specify sampling weights (via the Weights option from the Model menu or the WEIGhts command) was introduced as an experimental feature. Is this still the case, or has thorough testing now been completed?
- Can I use a title field to specify a chart title I want to see AND also use it to send CRChart Macro Commands?
- Why dont the svyset commands specify a finite population correction (fpc)?
