Converting ventiles to centiles

[sarahdykstra]

I have a dataset which provides the average income for people in each ventile, so twenty observations. I want to convert this to percentile data, with one observation for each percent of the population, assuming a normal distribution. Does anyone have suggestions for how to accomplish this?

[David Radwin]

I'm not sure how to implement it, nor whether it's even a good idea to impute centiles, but isn't a Pareto distribution more plausible for income?

[ben earnhart]

Sarah --

David is almost certainly correct. However, if you want an answer, try posting some data and info about the data (what ranges for what ventiles). It should be doable, but I'd need some data to play with and don't want to generate my own.

Read the FAQs about posting data: do it in a table, not as an Excel spreadsheet. And maybe we can work out a solution that isn't terrible.

At first blush, I thought it a silly thing to do. But if, for example, you have some countries in percentiles and others in ventiles, it's not so strange, at least for descriptives or computing Gini or something.

[ben earnhart]

I just realized that the range or shape of the distribution may not matter. If it really is ventiles, then it's a uniformly distributed, 20 point scale, right? In which case the code below is all you need. However, it gains you nothing, since those in a given ventile are randomly distributed amongst the centiles.

clear
set obs 1000
set seed 1971
gen x=rnormal()
xtile x20=x, nq(20)
xtile x100fromx20=x20, nq(100)