# Which test for significance: Compare counts between two time points [Excel]

#### sjmcd13

##### New Member
Hello -

First time poster. Thank you in advance for any advice that you can give.

I am conducting a simple descriptive analysis of patients who came in for a particular service over a 10-year period. I made the statement to my boss that demand for this service increased by about 450% over time by simply comparing the # of patient who came in during the first 12 mos (N=12) and the # who came in during the last 12 mos (N=67). When I told her this, my boss keenly asked if I *knew* whether this was a difference (significant) or not. I did not actually know for sure.

Basically, I want to confidently make the statement that demand increased over time but I'm not sure what test to use to check for significance. I have chosen to just compare the count of patients for the first 12 mos and the count of patients for the last 12 mos, with the implicit expectation that the # of patients per year would not change. What test would be best for this simple analysis?

At first I thought independent, two-tailed t-test. There is one population for the study, but this question is asking about the difference between two groups of diff people: those who came in during the first 12 mos and those who came in during the last 12 mos. But t-test isn't appropriate because I am not using means or
proportions.

Then I thought Poisson. But Poisson is really only appropriate for large counts that will conform to the Poisson distribution and these are not large counts. I also thought Chi2 for a min, but this is not categorical data and I don't really know what my expected amounts would be. I suppose I would expect that the # of people coming in
would be the same during the last year as the first, but that test still doesn't seem like a good fit.

I have checked the web and old stat texts. I feel like I am missing something really obvious. But I have not found the right test to answer this question: Test for significance comparing two counts between time 1 and time 2.

#### Karabiner

##### TS Contributor
You can use a Chi² test with 2 cells and the Null hypothesis
that the total n=79 patients come from a population with an
even (50%/50%) distribution across the two years inspected.

With kind regards

K.

#### sjmcd13

##### New Member
Thanks, Karabiner! This worked perfectly and jogged my rusty stat memory. Appreciate your help.

#### Jose Moreira

##### New Member
Hi, i have a doubt that i would like to clarify:

My question is:

How i calculate the weighted mean?

Ps.: see the file attached.

#### Jose Moreira

##### New Member
I sent you an attachement with an question that i want to be solved. And i´ll be available to support you with your doubts.