Dear research community, I am using a data file from DHS survey and I need to adjust for cluster sampling design. I am using SPSS version 27. When I weight the data based on the weights that have been provided by DHS my sample size changes from 4487 to 4620 and I don't even have 4620 individuals in the file. I do appreciate any solution or suggestion.
Because weighting means that some cases get counted as less than one. That's why you divide by sum of the weights not by n. Best D. Booth
Hello Shirin,
Professor Booth is correct; weighting means that an individual case or record in your data set can be counted as something different than 1.0 cases (could be more; could be fewer, depending on the furnished weights).
The "4620" value represents what your N would have been had all of the respective subgroups been proportionally represented, relative to whatever subgroups have a weight of 1. Think of it as a virtual N, if you like. If you require that the sum of weights = actual N for your data set, then multiply the given weights by the fraction (4487 / 4620) and re-apply the values.
Good luck with your work.
Dear Professor Booth and David, thank you so much for your excellent answers. My follow up question is that in the paper when we talk about the participants background characteristics (usually table 1), should we report actual N or weighted N or both?
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design