Random samples of data points. Each observation needs to be independent from one another (you can't have a single participant completing the survey more than once and there should be no interference between participants). Data being analysed needs to be categorical - therefore, no need for tests of normality or Levene's test of homogeneity of variance!