Chi-Squared Test

[Moose352]

What should I do when the expected value for a chi squared test is zero, so when I try to calculate the test statistic, i get a divide by zero?

[Moonbear]

The chi-squared test isn't very accurate when expected frequencies are less than 5 (and definitely not if they are 0). In some cases, you can instead use a Fisher's Exact Probability Test (if you have a 2x2 contingency table and small expected frequencies...I think less than 10). I don't have the formula for that handy though. If that doesn't work for your data, I'm not sure what other alternatives there are.