Background: I have two datasets with income estimates, one for the GLOBAL population, one for a LOCAL population (LOCAL only a small fraction of GLOBAL, on the order of 1/100).

Datapoints: I have only certain percentile values, mostly deciles (10th, 20th, 25th, 30th, 40th, 50th, 60th, 70th, 75th, 80th, 90th*). I also have the mean for both populations. [*90th not for LOCAL]

Goal: I'm interested in learning what fraction of the LOCAL population falls into each GLOBAL "decile band", so to speak, i.e. what fraction of the LOCAL population falls in between each of the NATIONAL percentile values that I do have.

Suggested trivial strategy: draw a straight line between each of the LOCAL percentile datapoints that I do have, and interpolate what the LOCAL 11th percentile is, and the 12th, and the 13th, and so on. Matching up to GLOBAL cutoffs will be trivial.

Additional info: More data might be available in past years estimates, or in more detailed breakdowns (income by gender, by employment status, + the proportion of population in each category, e.g. percent male etc.). I cannot access the raw data, only summary statistics such as this.

QUESTION: Is there a strategy likely to yield more accurate results than my suggested trivial strategy above? Is it worth trying to build a distribution of some sort rather than drawing straight lines? If so, how to know which one?



1 Answer 1


Try the R package rriskDistributions which includes a function 'fit.perc' which "provides a GUI for choosing a most appropriate continuous distribution for known quantiles." All you need to do is supply a vector of percentiles and a vector of corresponding quantiles.

  • $\begingroup$ Thanks-- this is the best solution I have found as well. $\endgroup$
    – W.S.G.
    Commented Jul 18, 2018 at 17:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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