Stata 18 dtable command - can I display single row for binary variables?

[Suzanna Vidmar]

I've just installed Stata 18 and have been trying out a few things with the new -dtable- command. I'm sure I'll use this command frequently in my work and especially like the ease of automating and reproducing descriptive tables. Thanks Stata!!!
I do however have one question/wish.

Here's my code:

*Reading in the low birthweight dataset available on the Stata Press website.
use https://www.stata-press.com/data/r18/lbw,clear

*Adding value label to the low birthweight and hypertension variables.
lab def noyes 0 No 1 Yes
lab val low ht noyes

*Creating descriptive table.
dtable age i.race i.ht, by(low, nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd)  

and the resulting output:

----------------------------------------------
                             Birthweight<2500g
                                No       Yes  
                              n=130     n=59  
----------------------------------------------
Age of mother                24 (5.6) 22 (4.5)
Race                                          
  White                      73 (56%) 23 (39%)
  Black                      15 (12%) 11 (19%)
  Other                      42 (32%) 25 (42%)
Has history of hypertension                   
  No                        125 (96%) 52 (88%)
  Yes                          5 (4%)  7 (12%)
----------------------------------------------

This is great when there aren't many variables. However, if there are many more variables, I prefer to display a single row for the binary variables, i.e. a single row for "History of hypertension" rather than 3 rows as shown above. I've attempted a work-around. Apart from only partially giving me what I want, it has introduced an undesirable consequence.

In my work-around I'm treating history of hypertension as a continuous variable (mean=percentage).

*ht is a binary 0/1 variable. Multiplying by 100 so its mean is the same as percentage.
gen ht100=ht*100
lab var ht100 "History of hypertension"

*Creating descriptive table.
dtable age i.race ht100, by(low, nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd) /// 
define(htperc=mean) sformat("(%s%%)" htperc) ///
continuous(ht100, statistic(htperc))

The resulting output is:

-----------------------------------------------------
                              Birthweight<2500g      
                              No             Yes     
                             n=130          n=59     
-----------------------------------------------------
Age of mother           24 (5.6) (24%) 22 (4.5) (22%)
Race                                                 
  White                       73 (56%)       23 (39%)
  Black                       15 (12%)       11 (19%)
  Other                       42 (32%)       25 (42%)
History of hypertension         4 (4%)       12 (12%)
-----------------------------------------------------

My questions are:

1. The percentage is repeated, first just the number and then in the specified format (#%). This is the case for both history of hypertension and mum's age (both treated as continuous). How do I tell Stata to display the mean only once within a cell? Can this be a different format for mum's age (showing just the mean, i.e. without (%)) vs history of hypertension (showing the mean with (%)).
2. Is there a way to show the frequency of people with a history of hypertension?

This is the output I would like to be able to produce:

----------------------------------------------
                             Birthweight<2500g      
                                No       Yes     
                              n=130     n=59     
----------------------------------------------
Age of mother                24 (5.6) 22 (4.5) 
Race                                                 
  White                      73 (56%) 23 (39%)
  Black                      15 (12%) 11 (19%)
  Other                      42 (32%) 25 (42%)
History of hypertension        5 (4%)  7 (12%)
----------------------------------------------

I'm aware my attempted work-around is convoluted. There may be a more straightforward way to do what I want that I missed when reading the help for -dtable-, in which case I will feel very silly.

Kind regards,
Suzanna

[Jeff Pitblado (StataCorp)]

You can restrict factor variable levels. In your call to dtable, use 1.ht instead of i.ht.
Then you can use collect style header to hide the factor's level in the row header with

collect style header ht, level(hide)
collect preview

Here is how I modified the original example

*Reading in the low birthweight dataset available on the Stata Press website.
use https://www.stata-press.com/data/r18/lbw,clear

*Adding value label to the low birthweight and hypertension variables.
lab def noyes 0 No 1 Yes
lab val low ht noyes

*Creating descriptive table.
dtable age i.race 1.ht, by(low, nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd)

collect style header ht, level(hide)
collect preview

And here is the resulting table

. collect preview

---------------------------------------------
                            Birthweight<2500g
                               No       Yes  
                              n=130    n=59  
---------------------------------------------
Age of mother               24 (5.6) 22 (4.5)
Race                                         
  White                     73 (56%) 23 (39%)
  Black                     15 (12%) 11 (19%)
  Other                     42 (32%) 25 (42%)
Has history of hypertension   5 (4%)  7 (12%)
---------------------------------------------

[Suzanna Vidmar]

Thanks Jeff for your quick and clear response. That's fabulous!

[amandacpac]

I would like to do this, but to also show the p-value for the test comparing the two groups. However, when I add the tests option to the by() statement, the table shows the 0 row. Here is the code I am using,

*Reading in the low birthweight dataset available on the Stata Press website.
use https://www.stata-press.com/data/r18/lbw,clear

*Adding value label to the low birthweight and hypertension variables.
lab def noyes 0 No 1 Yes
lab val low ht noyes

*Creating descriptive table.
dtable age i.race i.ht, by(low, tests nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd) ///
factor(ht, stat(fvfreq fvperc) test(pearson))


collect style header ht, level(hide)
collect preview

and the output

. collect preview

----------------------------------------------------
                                Birthweight<2500g   
                                No       Yes    Test
                              n=130     n=59        
----------------------------------------------------
Age of mother                24 (5.6) 22 (4.5) 0.103
Race                                                
  White                      73 (56%) 23 (39%) 0.082
  Black                      15 (12%) 11 (19%)      
  Other                      42 (32%) 25 (42%)      
Has history of hypertension 125 (96%) 52 (88%) 0.036
                               5 (4%)  7 (12%)      
----------------------------------------------------

[Jeff Pitblado (StataCorp)]

You need to specify the level of ht you want to show. The spots to change are

dtable age i.race i.ht, by(low, tests nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd) ///
factor(ht, stat(fvfreq fvperc) test(pearson))

Since the factor default statistics are fvfrequency and fvpercent and test is pearson, you can omit option factor() and just specify the level for factor ht.

dtable age i.race 1.ht, by(low, tests nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd)

Note that I changed the i-dot notation i.ht to a one-dot 1.ht.
Here is the resulting table

---------------------------------------------------
                               Birthweight<2500g   
                               No       Yes    Test
                              n=130    n=59        
---------------------------------------------------
Age of mother               24 (5.6) 22 (4.5) 0.103
Race                                               
  White                     73 (56%) 23 (39%) 0.082
  Black                     15 (12%) 11 (19%)      
  Other                     42 (32%) 25 (42%)      
Has history of hypertension   5 (4%)  7 (12%) 0.036
---------------------------------------------------

[amandacpac]

So, that i.ht was a typo. I fixed it in my code but apparently not in my comment. I get the same table that I posted above when I leave the factor() option in (with 1.ht). I get that I can omit that here, but if I use fishers exact or some other non-default, and thus need to include the factor() option, it seems to be messing it up.

[Jeff Pitblado (StataCorp)]

Here is how I changed your example, keeping option factor().

*Reading in the low birthweight dataset available on the Stata Press website.
use https://www.stata-press.com/data/r18/lbw,clear

*Adding value label to the low birthweight and hypertension variables.
lab def noyes 0 No 1 Yes
lab val low ht noyes

*Creating descriptive table.
dtable age i.race 1.ht, by(low, tests nototals) ///
sample(, statistic(frequency) place(seplabels)) ///
sformat("n=%s" frequency) ///
nformat(%2.0f fvpercent mean) nformat(%3.1f sd) ///
factor(1.ht, stat(fvfreq fvperc) test(pearson))


collect style header ht, level(hide)
collect preview

Here is the resulting table

---------------------------------------------------
                               Birthweight<2500g   
                               No       Yes    Test
                              n=130    n=59        
---------------------------------------------------
Age of mother               24 (5.6) 22 (4.5) 0.103
Race                                               
  White                     73 (56%) 23 (39%) 0.082
  Black                     15 (12%) 11 (19%)      
  Other                     42 (32%) 25 (42%)      
Has history of hypertension   5 (4%)  7 (12%) 0.036
---------------------------------------------------

[amandacpac]

Ahhh. That makes total sense. Thank you!