dtable and Removing Binary Labels in the Variable Column

[Kevin Blaine]

Hi Jeff Pitblado (StataCorp) -- is there a way to tweak the displaying of labels for the dtable command such that a binary variable's labels don't show and only a single value (i.e. "1" or "Yes") is reported on the same line as the variable name? Something like this:

[INDENT=3] [/INDENT]  
Variable
n (%)

Complex Chronic Condition Flag


 			
Neurologic Flag
34 (2.7%)

 			
Congenital Malformation Flag
88 (10.2%)

 			
Malignancy Flag
150 (12.2%)

Where the n's and %'s report the percent of cases where each variable = "1" or "Yes".

Does that make sense?

Thanks!

[Jeff Pitblado (StataCorp)]

You can select/restrict the factor levels shown by dtable by specifying them in the varlist.
For example,

sysuse auto
dtable 1.foreign

which yields table

---------------------
             Summary 
---------------------
N                  74
Car origin           
  Foreign  22 (29.7%)
---------------------

Then you can change the header style for factor variable foreign to hide it's level.

. collect style header foreign, level(hide)

. collect preview

---------------------
             Summary 
---------------------
N                  74
Car origin 22 (29.7%)
---------------------

[Kevin Blaine]

Brilliant! I suppose I'll have a lot of variables that this will be the case for. Can I specify something to the effect of: "Do that for all binary variables in the variable column."?

[Jeff Pitblado (StataCorp)]

There is no such option right now.
You can specify all the variables in a single call to collect style header though.

. dtable 1.foreign 5.rep78

-----------------------------
                     Summary 
-----------------------------
N                          74
Car origin                   
  Foreign          22 (29.7%)
Repair record 1978           
  5                11 (15.9%)
-----------------------------

. collect style header foreign rep78, level(hide)

. collect preview

-----------------------------
                     Summary 
-----------------------------
N                          74
Car origin         22 (29.7%)
Repair record 1978 11 (15.9%)
-----------------------------

[Kevin Blaine]

That'll do. Thanks for all the help!

P.S. You know an awful lot about dtable. Did you develop it?

[Kevin Blaine]

One more thing -- how do I APPLY a collected preview style? Do I have to save it as a new template, then call the dtable command again?

[Jeff Pitblado (StataCorp)]

I'm not sure what you mean by "collected preview style".

If you have a style that you saved to disk via collect style save, like I showed in the other thread, then you can apply it to your current dtable collection using collect style use filename, override. In the other thread I also mention specifying it in option style(). Unfortunately, header style edits on factor variables applied using option style() will be overwritten by code in dtable. The only way to apply header style edits to factor variables, currently, is to apply the style edits after the collection is created by dtable.

Here is an example. First let's define a style edit to hide the levels of factor variables foreign and rep78.

collect clear
collect style header foreign rep78, level(hide)
collect style save hide-levels, replace

Now let's apply this style after our call to dtable.

sysuse auto
dtable 1.foreign 5.rep78
collect style use hide-levels, override
collect preview

Here is the resulting table

. collect preview

-------------------------------
                   |    Summary
-------------------+-----------
N                  |         74
Car origin         | 22 (29.7%)
Repair record 1978 | 11 (15.9%)
-------------------------------

For this kind of customization, I would consider writing my own wrap-around program that identifies the 0/1 factor variables, calls dtable, then applies the header style edits.
Something like

program mydtable
    version 18
    syntax [varlist] [if] [in] [, indicators(varlist) *]
    quietly dtable `varlist' 1.(`indicators') `if' `in', `options'
    if "`indicators'" != "" {
        collect style header `indicators', level(hide)
    }
    collect preview
end

Here is the code in action

. mydtable mpg turn, indicators(foreign rep78)

---------------------------------
                       Summary   
---------------------------------
N                              74
Mileage (mpg)      21.297 (5.786)
Turn circle (ft.)  39.649 (4.399)
Car origin             22 (29.7%)
Repair record 1978       2 (2.9%)
---------------------------------

[Kevin Blaine]

Apologies for the confusion. First, all the detail is unbelievably helpful. Thanks!

With the override function, does it only override the aspect of the old style that the new style modifies? For example, per the previous thread, I've set up my ideal Table 1 style. (It's pretty great, IMHO!), but I don't want any additional style I create solely for the single-line variable label edits to overwrite the rest of the style. Does that make sense?

I think the wrap-around program idea is really solid, and essentially automates what I'm trying to do. Where I get a bit lost though (and the reason for my previous question about "applying" a collected style) is when I've modified a style using collect style [row/cell/etc], I no longer know how to export the table in the way that I would expect to using dtable's "export()" option. Even in the case of the wrap-around program, I still wouldn't know what to do re: exporting of the table to Word. Any insight on that too?

Thanks again for all the support,
Kevin

[Jeff Pitblado (StataCorp)]

Use collect export to publish your table into a specific file format. dtable is a convenience command, and its export() option is just a shortcut to call collect export.

[Jeff Pitblado (StataCorp)]

Regarding option override, it depends on what you started with when you defined your style edits.

Let's say we are working with dtable, and make style edits. Then override is not necessary because the style we save is already based on dtable. This is the situation I was proposing in this thread.

On the other hand, you could start from an empty style, via

collect style clear

make edits to that style, and save those edits to a file. If you specify this new file in option style(), you also want to add sub-option override, since you want all the default style choices we already made for dtable, but want your edits to be applied post hoc. If you fail to specify sub-option override with this file in option style(), then dtable would use the empty style choices followed by your edits.

The empty style basically means no colors, no borders, numeric format %9.0g, ..., etc. If you are curious about the style choices in the empty style, type the above command to clear the style, then follow it up with one of the following collect query commands.

collect query column
collect query row
collect query table
collect query header
collect query html
collect query putdocx
collect query putpdf
collect query tex
collect query showbase
collect query showempty
collect query show omit
collect query cell
collect query _cons
collect query stars
collect query title
collect query notes
collect query composite

[Kevin Blaine]

Thanks, Jeff. This answers my question perfectly. Much appreciated!

[Kevin Blaine]

Hi Jeff -- I'm back working on custom tables, and I'm wondering how I would show stars next to significant p-values in dtable? I've tried a bunch of different tacks, and none are taking. Ideas?

[Jeff Pitblado (StataCorp)]

dtable was not designed with stars in mind, but it is not too hard to replace the reported p-values once you figure out which result levels contain the p-values.

Here is an example

sysuse auto

dtable price mpg head trunk turn i.rep, by(foreign, test)
* -dtable- sets result autolevels to its composite results
collect query autolevels result
* the p-values are in composite _dtable_test: pearson and regress
collect query composite _dtable_test
* define stars rules using the p-value results
collect stars pearson regress .1 "*" .01 "**" .001 "***"
* redefine composite _dtable_test to use the new -stars- result
collect composite define _dtable_test = stars, replace
* replay table, now with stars
collect preview

Here is the resulting table, with stars

--------------------------------------------------------------------------------------------
                                                    Car origin                              
                             Domestic              Foreign                Total         Test
--------------------------------------------------------------------------------------------
N                                52 (70.3%)            22 (29.7%)           74 (100.0%)     
Price                 6,072.423 (3,097.104) 6,384.682 (2,621.915) 6,165.257 (2,949.496)     
Mileage (mpg)                19.827 (4.743)        24.773 (6.611)        21.297 (5.786)  ***
Headroom (in.)                3.154 (0.916)         2.614 (0.486)         2.993 (0.846)    *
Trunk space (cu. ft.)        14.750 (4.306)        11.409 (3.217)        13.757 (4.277)   **
Turn circle (ft.)            41.442 (3.968)        35.409 (1.501)        39.649 (4.399)  ***
Repair record 1978                                                                          
  1                                2 (4.2%)              0 (0.0%)              2 (2.9%)  ***
  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%)     
--------------------------------------------------------------------------------------------

[Kevin Blaine]

Thanks for the quick reply, Jeff. So, I stumbled upon that as a potential option, but I couldn't piece together how to report both raw p-values AND stars too. Any suggestions on how I could do that, using dtable as the primary command?

[Jeff Pitblado (StataCorp)]

In that case, instead of redefining composite result _dtable_test, add stars to the result auto levels.

dtable price mpg head trunk turn i.rep, by(foreign, test)
* -dtable- sets result autolevels to its composite results
collect query autolevels result
* the p-values are in composite _dtable_test: pearson and regress
collect query composite _dtable_test
* define stars rules using the p-value results
collect stars pearson regress .1 "*" .01 "**" .001 "***"
* add stars to the result autolevels
collect style autolevels result stars
collect style cell result[stars], halign(left)
* replay table, now with stars
collect preview

Here is the resulting table

--------------------------------------------------------------------------------------------------
                                                       Car origin                                 
                             Domestic              Foreign                Total            Test   
--------------------------------------------------------------------------------------------------
N                                52 (70.3%)            22 (29.7%)           74 (100.0%)           
Price                 6,072.423 (3,097.104) 6,384.682 (2,621.915) 6,165.257 (2,949.496)  0.680    
Mileage (mpg)                19.827 (4.743)        24.773 (6.611)        21.297 (5.786) <0.001 ***
Headroom (in.)                3.154 (0.916)         2.614 (0.486)         2.993 (0.846)  0.011 *  
Trunk space (cu. ft.)        14.750 (4.306)        11.409 (3.217)        13.757 (4.277)  0.002 ** 
Turn circle (ft.)            41.442 (3.968)        35.409 (1.501)        39.649 (4.399) <0.001 ***
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%)           
--------------------------------------------------------------------------------------------------