Question

Help with Chi Squared !! I have the following data set:

Watersheds: 0 1 2 3
Autism: 10 4 1 0 15
Total: 213 146 25 24 408

The data represents search and rescue cases and how many watersheds away people are usually found. So an individual could be found within the same watershed (0) an adjacent watershed (1) and so on. The total column/row represents how many cases are found for instance 213 cases out of 408 are found in the same watershed and so on.

What I am trying to do is perform a chi squared test to see if the Autism distribution matches the overall (total) distribution.

Answer 1

Chi-square "goodness of fit" tests compare your count data against what is predicted by your hypothetical distribution. Your hypothesis is that the data on autism cases matches the general population, you have to compare the autism counts against what would be expected if they were in the same proportion as the overall population.
watershed distance = 0 would be expected in (214/408)(15)= 7.9, so the first contribution to your chi-square total would be (10-7.9)^2 / 7.9

Answer 2

Thank You very much for your help! I was also hoping if you could please help me with one more problem. I am supposed to do a T-test to see if model C is statistically better than model A. Below are the results of the models:
Model A has an average score of 0.781 (95% CI: 0.774-0.818)
Model B has an average score of 0.618 (95% CI: 0.566-0.64)
Model C has an average score of 0.809 (95% CI: 0.779-0.847)

How would I go about doing such test?

Answer 3

While you could calculate a t statistic, based on the difference between the means and a pooled estimate for the standard error of the means, the fact that the 95% C.I.s overlap a lot indicates the means are not statistically different. Since no # samples is given, I assume the C.I.s are for the means rather than for the populations.

Answer 4

The samples for each model are as follows:
Model A = 398 Model B= 392 and Model C= 392

How would I go about doing the test with these samples?