Chi-square or test of equal proportions?


First post on the forum. I'm currently studying for my ordinary certificate from the Royal Statistical Society and using the textbook "Introducing Statistics" by Upton and Cook.

Dispite pouring over the textbook I'm still confused as to which test to use for the following situation which has come up at work.

A hospital wanted to improve its mortality rates for a particular disease and implemented a new procedure. In the year before the new procedure, 1,139 patients died while 9,476 survived. A year after implementing the new procedure, 684 patients died while 10,698 survived. Was the new procedure a success or could these figures be down to chance variation?

So the hypothesis is that there is a significant lowering of deaths following the implementation of the new procedure.

I'm wondering if I should use a chi-square test for this or a test of equal proportions. To be honest I don't fully understand the difference. But I'm assuming that which ever test I use, with 95% confidence, if the p-value is less than 5% it is highly unlikely that the difference was due to random varianation and thus it can be inferred that the new procedure has indeed had a significantly beneficial effect on the survival rate.

So is it a chi-square or equal proportions test or something else I haven't even thought of?

I'm using R by the way.

Thanks Dason for the reply.

I did both at work and got the same result p = 2.2e-16. Then again I get that result a lot in R.

I've been reading tonight Statistics for People Who Think They Hate Statistics by Salkind and he mentions the McNemar Chi Square "to examine before and after changes." I put the data into R just now but get the old 2.2e-16.

So it looks like the p value is indeed very small no matter the test. So does this mean, all other things being equal, the intervention had a positive result on mortality levels i.e. they reduced?


Edit: Acutally it seems that McNemars test is for dependent variables whereas in this test they are independent.
Last edited: