I have a dataset where I have respondents from 3 different countries (USA, NL & DE) and use the Kruskal-Wallis to determine if there are any statistically significant differences regarding questions that were asked in a 7-point Likert item format.

The problem is that there is a specific variable (sampling method), where the three groups have very different rates in (USA 80-20%, NL 80-20%, DE 50-50%). In order to cancel out any effects that the sampling method could have, I was planning on weighing cases so that each nationality has the same rate. Would that be possible and if not, what other options do I have?

  • $\begingroup$ Describe the sampling designs in detail. The ordinary KW test requires independent observations, If the designs had a stage where clusters of respondents (e.g. schools), then the ordinary test cannot be used. Also, a sampling rate refers to a fixed proportion which was chosen by design to be sampled. It does not refer to the responding fraction of a larger sample. $\endgroup$ – Steve Samuels May 18 '15 at 22:16
  • $\begingroup$ Also what are the "rates" you refer to? $\endgroup$ – Steve Samuels Jun 2 '15 at 2:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy