Category-wise comparisons in Stata 18’s dtable

Stata 18’s new dtable is terrific for creating a “Table 1”, but by default if provided a categorical variable and asked to test for differences by group, it will provide a single Chi-sq test by default.

. sysuse auto, clear
(1978 automobile data)

. dtable i.rep78, by(foreign, tests)
note: using test pearson across levels of foreign for rep78.

-----------------------------------------------------------
                                  Car origin
                    Domestic    Foreign     Total     Test
-----------------------------------------------------------
N                  52 (70.3%) 22 (29.7%) 74 (100.0%)
Repair record 1978
  1                  2 (4.2%)   0 (0.0%)    2 (2.9%) <0.001
  2                 8 (16.7%)   0 (0.0%)   8 (11.6%)
  3                27 (56.2%)  3 (14.3%)  30 (43.5%)
  4                 9 (18.8%)  9 (42.9%)  18 (26.1%)
  5                  2 (4.2%)  9 (42.9%)  11 (15.9%)
-----------------------------------------------------------

Instead you may want a category-by-category comparison of proportions. We’ll need to create a series of binary variables first.

. quietly tab rep78, gen(repcat)

. label var repcat1 "Repair Record Category 1"

. label var repcat2 "Repair Record Category 2"

. label var repcat3 "Repair Record Category 3"

. label var repcat4 "Repair Record Category 4"

. label var repcat5 "Repair Record Category 5"

Next, we can add each of the repcat* to dtable, but preface with 1. to only include the “yes” for each binary.


. dtable 1.repcat*, by(foreign, tests)
note: using test pearson across levels of foreign for repcat1, repcat2, repcat3, repcat4, and repcat5.

-----------------------------------------------------------------
                                        Car origin
                          Domestic    Foreign     Total     Test
-----------------------------------------------------------------
N                        52 (70.3%) 22 (29.7%) 74 (100.0%)
Repair Record Category 1
  1                        2 (4.2%)   0 (0.0%)    2 (2.9%)  0.342
Repair Record Category 2
  1                       8 (16.7%)   0 (0.0%)   8 (11.6%)  0.047
Repair Record Category 3
  1                      27 (56.2%)  3 (14.3%)  30 (43.5%)  0.001
Repair Record Category 4
  1                       9 (18.8%)  9 (42.9%)  18 (26.1%)  0.036
Repair Record Category 5
  1                        2 (4.2%)  9 (42.9%)  11 (15.9%) <0.001
-----------------------------------------------------------------

So we have the tests we want (at least, if you’re OK with a Chi-sq test for each binary against the group variable), but the extra “1” entries is redundant. We can use collect to modify this.

. collect style header repcat1 repcat2 repcat3 repcat4 repcat5, level(hide)

. collect preview

-----------------------------------------------------------------
                                        Car origin
                          Domestic    Foreign     Total     Test
-----------------------------------------------------------------
N                        52 (70.3%) 22 (29.7%) 74 (100.0%)
Repair Record Category 1   2 (4.2%)   0 (0.0%)    2 (2.9%)  0.342
Repair Record Category 2  8 (16.7%)   0 (0.0%)   8 (11.6%)  0.047
Repair Record Category 3 27 (56.2%)  3 (14.3%)  30 (43.5%)  0.001
Repair Record Category 4  9 (18.8%)  9 (42.9%)  18 (26.1%)  0.036
Repair Record Category 5   2 (4.2%)  9 (42.9%)  11 (15.9%) <0.001
-----------------------------------------------------------------
Home | Back to blog