I have data from a survey that was (attempted to to be) administered to all children of particular grades in a certain state. I am getting it after a cleaning step by the survey designers that removed obvious invalid answers (from obnoxious teens).
Questions A and B have binary answers, and I'm interested in reporting the percentages of children in the 2x2 categories.
Of the approximately 100,000 observations I have, there are 3,500 that have missing data for either A or B, and are not included in the table. There's decent reason to believe that for these 2 questions, non-responses won't be particularly biased one way or the other.
What is the proper way to test/summarize any differences between the categories? Is a chi-square test meaningful here? If the non-responses are unbiased, do I just have a really large random sample? Or can I assume that I am just reporting the actual proportions and no statistical testing is needed?