Collapse a variable for every percentile

[Etienne Bergeron]

Hi,

I have a dataset of wages at the individual level. I want to collapse this dataset at every percentile.

I could simply code

collapse p(1) p1wage=wage p(2) p2wage=wage p(3) ... p(98) p98wage = wage p(99) p99wage = wage,

but I am looking for a more efficient way to do so, if it exist.

Thank you in advance,

Étienne

[Carole J. Wilson]

Have you considered something like:

pctile pc=wage, nq(100) gen(newp)
drop if mi(pc)
keep pc newp

[Joro Kolev]

It seems to me that the collapsing you use simply would produce the 1-100 percentiles of your data.

If this is what you want to calculate, the 1-100 percentiles of your data, a simpler way to obtain the calculation is :

centile wage , centile(1/100)

[Carole J. Wilson]

Note that pctile and centile will produce the same result only if you add the option -altdef- to pctile.

[Etienne Bergeron]

That is indeed what I'm looking for, thanks a lot!

[Joro Kolev]

Note that pctile and centile will produce the same result only if you add the option -altdef- to pctile.

Thank you for this very useful clarification, Carole !

I did notice that the results -centile- produces are slightly different from the results -summarize, detail- produces at few percentiles.