0
$\begingroup$

What is the approach to use to determine the statistical significance of change in incidence rate over years?

Below is at table showing such kind of data of incidence rate per 1000.

+------+---------+--------+
| Year | Malaria | Trauma |
+------+---------+--------+
| 2015 |  8.5    |0.2     |
| 2016 |  17.4   |0.2     |
| 2017 |  9.0    |0.0     |
| 2018 |  13.2   |0.3     |
| 2019 |  13.7   |0.3     |
+------+---------+--------+

Is there a formula or technique that will show whether the change in incidence rate between any two consecutive years and/or 2015 to 2019 is significant, given a predetermined P value of say 0.05

$\endgroup$
2
  • $\begingroup$ No, not with that data. At least not properly. You need total numbers. Moreover, 0.0 is probably meaningless, depending on the population size, single or zero significant figures are not very reliable when they likely have been overly rounded to the point of absurdity. $\endgroup$
    – Carl
    Mar 29, 2020 at 16:19
  • 1
    $\begingroup$ Why do you want to compare two years & why do you care about "statistical significance"? Different questions/goals require very different methods. Often interest is not just in "there is a difference between years that's not just chance variation under some assumed distribution" (easish), but in e.g. did some public health policy affect incidence (then simple methods for "Was something different?" are completely inappropriate. A more minor thing that makes it hard to answer is how you define incidence (e.g. can you clarify if a person can count twice or not - I'd normally assume the latter). $\endgroup$
    – Björn
    Oct 22, 2021 at 20:50

1 Answer 1

1
$\begingroup$

Yes there is but you need less aggregated data. You now only have 5 data points which is usually not enough for statistical testing. Could you maybe get the monthly or even daily data?

When comparing 2 years you can do a t-test on the daily data for example with incident rate as numerical and year as grouping variable.

When choosing to do a time series analyses you can try to predict incident rate with time using for example a regression analysis. If that is significant it means there is significant change over time in incidence rate.

However, for both you need more detailed data! If not possible you can just compare the differences without any statistical testing and use your common sense.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.