comparing prevalence bwteen quartiles

[abdelilah arredouani]

Dear all,
I have a continuous variable X that i categorized into 4 quartiles (1,2,3 and 4). i also have a binary variable Y (1=obese, 0=normal-weight). i can see from the tab command that obesity prevalence is lower in quartile 4 as compared to quartile 1 (table below). How do i go to test the prevalence of obesity between the quartiles or at least between Q1 and Q4.

Thanks in advance

Stat 15.1 Mac

[Carlo Lazzaro]

Abdelilah:
see Example 5 under -qreg- entry in Stata .pdf manual.

[abdelilah arredouani]

I just want to compare proportions between quartiles. i can't figure it out from the manual Carlo suggested. or maybe i am on the wrong track

[Carlo Lazzaro]

Abdelilah:
if you use your original (ie, continuous) X variable, you can do something along the following lines:

use https://www.stata-press.com/data/r16/auto, clear
. sqreg price foreign, q(.25 .5 .75) reps(100)
(fitting base model)

Bootstrap replications (100)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
..................................................    50
..................................................   100

Simultaneous quantile regression                    Number of obs =         74
  bootstrap(100) SEs                                .25 Pseudo R2 =     0.0063
                                                    .50 Pseudo R2 =     0.0219
                                                    .75 Pseudo R2 =     0.0118

------------------------------------------------------------------------------
             |              Bootstrap
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
q25          |
     foreign |        312   423.0919     0.74   0.463    -531.4184    1155.418
       _cons |       4187   135.4454    30.91   0.000     3916.995    4457.005
-------------+----------------------------------------------------------------
q50          |
     foreign |        983   611.1209     1.61   0.112    -235.2473    2201.247
       _cons |       4816   230.1752    20.92   0.000     4357.154    5274.846
-------------+----------------------------------------------------------------
q75          |
     foreign |        975    1673.69     0.58   0.562    -2361.439    4311.439
       _cons |       6165   993.2648     6.21   0.000     4184.963    8145.037
------------------------------------------------------------------------------

. test [q25]foreign = [q75]foreign

 ( 1)  [q25]foreign - [q75]foreign = 0

       F(  1,    72) =    0.16
            Prob > F =    0.6920

. lincom [q25]foreign - [q75]foreign

 ( 1)  [q25]foreign - [q75]foreign = 0

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         (1) |       -663   1666.858    -0.40   0.692     -3985.82     2659.82
------------------------------------------------------------------------------

.

[abdelilah arredouani]

that is complicated. What about prtest. it only works with 2 categories. is it possible to select Q1 and Q4 in prtest?

[Nick Cox]

Arguably a chi-square test ignores the ordering of your predictor's bins, but it surely qualifies as simple: Other tests respect the ordering.

. tabi 17 51 \ 21 38 \ 23 51 \ 42 24

           |          col
       row |         1          2 |     Total
-----------+----------------------+----------
         1 |        17         51 |        68 
         2 |        21         38 |        59 
         3 |        23         51 |        74 
         4 |        42         24 |        66 
-----------+----------------------+----------
     Total |       103        164 |       267 

          Pearson chi2(3) =  24.7579   Pr = 0.000

. ret li

scalars:
                  r(p) =  .0000173486835944
               r(chi2) =  24.75792545374551
                  r(c) =  2
                  r(r) =  4
                  r(N) =  267