Confidence intervals of medians by group

[Toni Moreno]

Dear all,

I want to find out the confidence intervals of the median earnings by different years.

Unfortunately, the centile command has no by-option so that something like:

centile earnings, by (year)

does not work.

I also thought about the or ci command, but as far as I'm aware it does not allow to calculate the confidence interval of medians.

Any help would be appreciated.

[Carlo Lazzaro]

Toni:
I do hope that what follows may be helpful:

. use "C:\Program Files (x86)\Stata15\ado\base\a\auto.dta"
(1978 Automobile Data)

. bysort foreign: centile price, centile(10(10)90)

------------------------------------------------------------------------------------------------------------------------
-> foreign = Domestic

                                                       -- Binom. Interp. --
    Variable |       Obs  Percentile    Centile        [95% Conf. Interval]
-------------+-------------------------------------------------------------
       price |        52         10      3866.8        3297.907    4087.846
             |                   20      4092.2         3890.42    4418.487
             |                   30      4369.6         4077.78    4618.593
             |                   40      4506.4        4190.473    4905.179
             |                   50      4782.5        4464.559    5316.421
             |                   60      5185.6         4729.55     6043.89
             |                   70      5806.8        5118.746     8016.34
             |                   80      7022.4        5789.575     11442.4
             |                   90     11463.4        7365.857    14692.03

------------------------------------------------------------------------------------------------------------------------
-> foreign = Foreign

                                                       -- Binom. Interp. --
    Variable |       Obs  Percentile    Centile        [95% Conf. Interval]
-------------+-------------------------------------------------------------
       price |        22         10      3827.1            3748    4467.642*
             |                   20      4175.6         3769.71    5027.452
             |                   30        4580        3904.144      5763.5
             |                   40      5142.6        4270.964    6194.331
             |                   50        5759         4582.73    6870.205
             |                   60        6163        5112.408    8258.839
             |                   70        6879          5754.5    9730.885
             |                   80      8753.4        6237.906    12557.97
             |                   90       11317        7292.772       12990*

* Lower (upper) confidence limit held at minimum (maximum) of sample

.

[Joro Kolev]

Where are you taking this information that you are reporting from ? The -help centile- says " by and statsby are allowed; see prefix."

Here:

. sysuse auto, clear
(1978 Automobile Data)

. bysort foreign: centile price, cen(50)

------------------------------------------------------------------------------------------------------------------
-> foreign = Domestic

                                                       -- Binom. Interp. --
    Variable |       Obs  Percentile    Centile        [95% Conf. Interval]
-------------+-------------------------------------------------------------
       price |        52         50      4782.5        4464.559    5316.421

------------------------------------------------------------------------------------------------------------------
-> foreign = Foreign

                                                       -- Binom. Interp. --
    Variable |       Obs  Percentile    Centile        [95% Conf. Interval]
-------------+-------------------------------------------------------------
       price |        22         50        5759         4582.73    6870.205

.

[Toni Moreno]

Thank you so much for your help and sorry for misreading the help-file.

[ericmelse]

Dear Toni,

Possibly, this paper offers good advice to compare and examine median values between group categories:

The Stata Journal 2012 12 2 162-190 RM Conroy What hypotheses do nonparametric two-group tests actually test?

Best,
Eric

[Marcos Almeida]

ericmelse Thank you for sharing the article in #5. Quite insightful!