I am working on sampling a posterior distribution for parameters in my model using approximate Bayesian computation (ABC). I would like to come up with a summary statistic to compare the similarity/distance of ranked order between my simulated data and observed data. My observed data are a list of 10 species biomass. So I would like to compare the rank of species i between simulated and observed data, and summarize over 10 species.

My questions are:

  1. Is there some metrics that can quantify the rank differences between 2 vectors? So far, what I did is to take the rank difference per species and calculate the sum of absolute difference across the 10 species.

  2. Because the metric should provide a distance-like result: the more similar the 2 vector ranks, the smaller the value is. I am not sure if I can use the test statstics in Wilcoxon signed-rank test. Anyone has any comments?

Thanks very much


Can you use a simple Spearman correlation? It will give you a rank-based correlations statistic with the benefit of having metric properties, I believe.

  • $\begingroup$ thanks very much. The Spearman rank correlation is exactly what I want $\endgroup$ – tiantianchen Aug 17 '20 at 7:46

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.