Question

You are given a chi square to fill out with 4 different observed values: 0, 54, 36, and 18. These numbers represent the 4 different phenotypes of offspring. If you have an observed value of 0 for one of these variables, can it even be included as a probability? Would that row just be disregarded from the data, and as a result you would have to adjust the degrees of freedom? (From 4 categories to 3, then - 1)

Had an exam question like this today and this was the best I could word my thinking. I couldn't find anything online with an example like this. I know with a similar question we had, when the expected value was 0, the row was disregarded, but I'm not sure if it's the same case for observed.

Answer 1

The row should not be disregarded. Just continue with the test as you normally do. Do not bother if an observed value is 0 or not. No need to adjust the degrees of freedom. You can not delete a category. It's only that in a certain sample, the observed frequency of such one category is 0 and it does not mean that the frequency of that category in that population is 0 (if it is so, then that category won't even exist. If they mentioned about that category, then it means it exists. So, don't disregard it and don't adjust the degrees of freedom).

If you have encountered with such a situation in your research process and not in an exam, then you need to study the population to check at least one subject of that category exists in the population where you got 0 frequency in the sample.