Recall that we wanted a way to test the equality of all of our proportions at once. But now that you have found "at least one proportions differs from the others," the natural follow-up question is which ones differ.

We won't get too far into the theory here, but one approach when you have rejected the null hypothesis is to calculate all pairwise confidence intervals to see which do not contain zero.

  • In the applet, with the Chi-square statistic selected, you can check a box to Compute 95% confidence intervals.
  • Include a screen capture of the confidence interval output in your report.

(q) Which intervals do not contain zero? What does that tell you about which population proportions differ from which other (and in which direction) and which do not?


(r) Describe how these intervals would differ if we had used 90% confidence instead of 95% confidence.



Back Next