Standard error of estimates’ difference in Complex overlapping samples

#1
Hi All,

I am dealing with survey data sourced by means of complex sample designs (involving stratification and clustering) across two survey waves.

The two samples are to some extent overlapping, as some participants from the first wave have also responded to the second year’s wave. And I need to estimate the standard error of the difference between estimates from the two years.

The standard errors in the second year should be partially accounted for by those in the first year. Therefore the formula [s12/n1+ s22/n2]^0.5 is probably not an accurate reflection of the standard error of the difference.

So far I have been calculating the standard error of the differences manually for all variables in my dataset (deflating it by a fraction of the standard error in the wave 2 sample), but I’m hoping there is a more clever way to tackle this automatically, avoiding the manual calculations.

Does anyone know if this is possible within SPSS Complex Samples (or possibly other solutions)?

Thanking you in advance