Percentile error

[samaneh khaef]

hi.
I am working with register data and I can't extract them to show you how it looks like. I try to explain clearly so. I have a simple population for thousands of neighborhoods. I sort the population across these neighborhoods. then I want to have percentile as this:
xtile p=population, nquantiles(100)
But when I tab I see the weird result as:
tab p

p	No
1	13000
47	12456
48	1700
56	1560
57	1498
58	1333
59	1200
85	1000
86	950
87	932
88	688
89	512
90	450
91	315
92	310
93	300
94	290
95	289
96	270
97	251
98	283
99	290
100	298

I am very new to percentile and I do not know where I made mistake. But I should get the equal frequency and also g from 1 to 100. why I just get some quartile and not with even distribution?

[Nick Cox]

No mistake in Stata terms, just a (frequent) misunderstanding, To get 100 bins with at least one value in each, it is necessary (but not sufficient) to have at least 100 distinct values. Further, values that are identical must end in the same bin (just as they do in any table or histogram). So ties can frustrate binning by quantiles. On the face of it your variable is utterly unsuitable for quantile binning.

I will post some references later.

[samaneh khaef]

Thanks for your reply! so how can I divide my population in percentile? indeed as I have a deprivation index, I need to make percentile and then plot percentile of the population with regard to deprivation. is there any way to do percentile for my poputaion?

[Nick Cox]

You can calculate a percentile rank. There is a FAQ on that on the Stata website, which

search percentile rank, faq

should find.

Why bin or even rank at all? That’s a way to throw away information.

It is difficult for me to know why even a rank is preferable in any sense to using a deprivation index itself (possibly on a transformed scale, as I gather its distribution is highly skewed).

[Nick Cox]

References

FAQ https://www.stata.com/support/faqs/s...ing-positions/

Papers on quantile binning, and especially why it is often a disappointment verging on disaster:

https://www.stata-journal.com/articl...article=pr0054 (PDF is free)

https://www.stata-journal.com/articl...article=dm0095

[samaneh khaef]

Thanks Nick!