Hello, everyone.
I’m working with two waves/years’ worth of data from a nationally representative survey, and I need to combine the two samples to work with a larger sample in a cross-sectional analysis. Each database has a final weight variable plus 100 replicated weights - the method used to calculate standard errors is the delete-a-group jackknife. They do not provide variables such as PSU and strata due to confidentiality concerns.
After using the `append` command to merge the datasets, I declared the dataset as a survey using this command (does this look correct?):
svyset [pw = finalweight], vce(jackknife) jkrweight(finalweightrep1 - finalweightrep100) mse
Seems ok to use the weight variables without adjustment to estimate proportions or use regression models?
Also, I need to create variables in quintiles (I think it is not correct create quintiles variables before the append, isn't it?). In this case, should I weight this variable? The `xtile` option has three types of weighting—which one should I use, considering the weighting method used in this research?
Thanks in advance!
I’m working with two waves/years’ worth of data from a nationally representative survey, and I need to combine the two samples to work with a larger sample in a cross-sectional analysis. Each database has a final weight variable plus 100 replicated weights - the method used to calculate standard errors is the delete-a-group jackknife. They do not provide variables such as PSU and strata due to confidentiality concerns.
After using the `append` command to merge the datasets, I declared the dataset as a survey using this command (does this look correct?):
svyset [pw = finalweight], vce(jackknife) jkrweight(finalweightrep1 - finalweightrep100) mse
Seems ok to use the weight variables without adjustment to estimate proportions or use regression models?
Also, I need to create variables in quintiles (I think it is not correct create quintiles variables before the append, isn't it?). In this case, should I weight this variable? The `xtile` option has three types of weighting—which one should I use, considering the weighting method used in this research?
Thanks in advance!
