The company where I work did a survey with a complex sample. Originally, the sample was a 2-step stratified, but due to some problems, we lost one of them (so, let's consider that the sample is stratified only on 1 variable).

The biggest part of the questions is like "What you think the company can do in respect of X?" and then there is a list of options (some questions allow only 1 answer, some more than 1, others are Likert-type response options).

My question is: Wow can I analyze this data? It would be nice if there were a test like Chi-square for complex sample (probably there exists one, but I don't know)

Any software is welcome, but I prefer R and SPSS.


1 Answer 1


There are lots of methods for dealing with complex survey data. Almost all of them require you to be able to specify the complex structure of the sampling framework.

In R there is the survey package (see CRAN repository for help files) by Thomas Lumley, who has also written a book on the topic "Complex surveys: A guide to data analysis in R", which is quite readable from a theoretical perspective and very useful from the applied perspective (so useful that my copy is currently sitting in a colleagues office.)

I haven't tried complex survey analysis in SPSS: you need an add-on to do anything that takes proper account of survey design (which will cost you money.)

But the R package should do everything you want it to do, for free, and with the greater flexibility that comes with R.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.