0
$\begingroup$

I have a dataset that is composed of homicide rates (per 100,000 or per 10,000) from 1980-2014 for 11 geographical areas. I need to assess whether there are statistically differing rates between the geographical areas over those 24 years. I am not certain how to approach this analysis or what would be the best way to go about testing for difference. Chi-squared testing has been suggested to me but I'm still unsure how to apply it to this dataset. Would you be able to suggest a way to assess for statistically significant differences between these geographical areas over the time specified? If possible, could you suggest ways to approach this using R software?

$\endgroup$
  • $\begingroup$ What is the specific hypothesis you want to test? A trend of change over the 24 year period, or a difference between the average rate over all 24 years, or something else? $\endgroup$ – Don Walpola Feb 25 at 2:38
  • $\begingroup$ I'm looking to establish whether a difference exists in any pairwise comparison of the areas over the 24 year span $\endgroup$ – Emily Feb 25 at 19:08
  • $\begingroup$ Then you would need a full multivariate model. $\endgroup$ – kjetil b halvorsen Feb 25 at 21:22
0
$\begingroup$

So you have a time series (or multiple time series, one for each area) of yearly counts. I would start out with Poisson regression, with $\log(10 000)$ or $\log(100 000)$ as an offset, so as to modeling rates. This is called Poisson rate regression, search this site. You must have an eye out for possible overdispersion.

As for the time series aspect, look through this earlier Qs with answer. As for using R, this post Modeling multivariate Time Series Count Data in R have jags code run from R, this post have some pure R code.

| cite | improve this answer | |
$\endgroup$

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.