Dear Alexandar,

Considering you would like to compare incidence rates (IR) in the same population across two years (Y1 vs Y2), i guess you are searching for a causal association between a certain risk factor (intervention on infection disease population) and a REPEATED measured outcome (i.e. to test intervention performance). Nevertheless, a PAIRED t-test would be useful to determine if the two sets of quantitative data (I.E. outcomes in the two separated data points, that is Y1 and Y2) are significantly different from each other (considering two measures of the same individual in different data points). That would explain if the risk factor or intervention has an effect on disease population. Importantly, data distributions when using a t-test have to be normally distributed. I don't know if your outcome is directly the IR of a different dependent measure but if so, I recommend you to use a standardized ratio, because I don't know if you can directly compare the IR with a t-test. I guess it would be interesting that you explain your aims (hypothesis) a bit more detailed.

Regards,

Dear Alexander,in order to give a specific answer, as Martin pointed out, it would be better if the the hypothesis to be tested is known

As far as i understand, the same sample before and after will act as control and experimental sample respectievely, and the study is aimed at testing the efficacy of the intervention (since u say before and after intervention) paired t-test could be employed

regards

Rathish

Dear Akexandra,

It would be better if you could put your research hypothesis to be tested.

However, in line with your query i believe you can use paired t-test if the data obtained shows normal distribution at different data points of same sample.

Dear Akexandra

i think  you incidence ratio to compare paired data 

data in this case represented qualitative type and i think we should use Mcknemar test non parametric test for paired qualitative data

Ok. Thanks all! ^_^

Hi Aleksandar P. Medarevic , I'm interested to know what test you decided to use to analyse your data? I have a similar problem - I want to compare the incidence rate of some behaviours in my animal subjects before and after an intervention, but can't find a suitable test for paired incidence rates. Any suggestions? :)

Dear Lorelle Barrett ,

I have the same problem now. I am searching how to test two dependent incidence rates, i.e. before and after an intervention in the same space. However, I couldn't find any answer, except considering Poisson distribution. Did you find any solution??

Hi Afra Alkan ,

In the end I tried using the Wilcoxin test, but my sample size is very low and the results were a bit suspect. This data has been hard to handle, and I have resorted to doing a descriptive analysis. I calculated a baseline rate of behaviour for each individual (mean and 95% confidence interval) before the intervention, then identified any rates of behaviour that lay outside of the individual's CI after the intervention. Then created a table that showed how many individuals had rates of behaviour that were below, within, or above their CI. Anything where the majority of individuals were either above or below their CI was considered a result of interest, but can't call it statistically significant.

Actually CI works well for inference, but here, researchers request a p-value, since they don't comprehend CI in general.

Thank you very much Lorelle Barrett for your response.

T test can be used for similar samples and not for independent samples so that you can extract the moral differences of variables you intend to study now ... Greetings to you