Tabulations and Tables

[Ghia Osseiran]

Simple question: tab x y,row nolabel
tab x z, row nolabel

What would be the command for a table that combines the above two tab commands?
So it includes x y z in row without the label

I tried table x y z, row
But for some reason that's missing a column ( i have the header for all four columns but it's missing the output of one column)

Thanks,
Ghia

[Nick Cox]

Not clear to me what your table would look like, but try

ssc inst groups 
help groups 

webuse nlswork
groups race collgrad south, sepby(race)

  +-------------------------------------------+
  | race   collgrad   south   Freq.   Percent |
  |-------------------------------------------|
  |    1          0       0   10886     38.16 |
  |    1          0       1    5442     19.08 |
  |    1          1       0    2670      9.36 |
  |    1          1       1    1176      4.12 |
  |-------------------------------------------|
  |    2          0       0    2689      9.43 |
  |    2          0       1    4495     15.76 |
  |    2          1       0     331      1.16 |
  |    2          1       1     534      1.87 |
  |-------------------------------------------|
  |    3          0       0     194      0.68 |
  |    3          0       1      27      0.09 |
  |    3          1       0      73      0.26 |
  |    3          1       1       9      0.03 |
  +-------------------------------------------+

[Ghia Osseiran]

Thanks this worked but what I'm looking for is more a table with race unigrad (dummy variable so 0 1) and say vocational grad ( 0 1) where I just get the frequency of each per race in this example. So the table would be five columns in total (exactly what you get from the command tab x y, row just with 3 variables instead of 2).

[Nick Cox]

Check out tab2, firstonly

[eric_a_booth]

Hi Ghia: I think some example data might make your question more clear, but here are some thoughts:

1. You mention: "I tried table x y z, row
But for some reason that's missing a column ( i have the header for all four columns but it's missing the output of one column)"

This is due to missing combinations of xy, yz, ... in your table -- where those combinations don't exist, -table- leaves the column blank in the output.


2. Your original post says you want the output from -tab x y,row nolabel- combined with the output from -tab x z, row nolabel- . Importantly, this produces row percentages in the -tab- output in addition to the frequencies (later you talk about just frequencies in lieu of row percents). If you want to use -table- to produce row percents (though there are better ways to get there, see below), you'd need to manipulate your data so that that the value produced in the -table- are the row percents. One way to do that is to use -contract- like this:

sysuse auto , clear
keep if mpg<20
tab mpg rep78,row nolabel
tab mpg for,row nolabel

**
table mpg rep78 for


preserve
    contract  mpg for rep78,  
    bys mpg: egen denom = total(_freq)
    g rowpct = round((_freq/denom)*100, .01)

 table mpg rep78 for , c(mean _freq) //same as prior table cmd
 table mpg rep78 for, c(mean rowpct) //now w/row pcts
restore

however, based on your later response to Nick, I'm suspect you might want frequencies only. If so, -table- is already giving you what you need (without any data manipulation like contract).

**Producing 3-way tabulation options:
You can produce 3 way tabulations by using something like -contact- or -collapse- and then export your dataset. You can also download -tab3way- from SSC.

However, I prefer to combine the categories for 2 of the variables, and then produce the tabulate table (or -tabout- from SSC is another good option here) using the new combined variable.

__
Here is an example using fake data that mirror what you have described about your dataset:

**create fake data
clear
set obs 100
g unigrade = rbinomial(1, .4)
g vocational = rbinomial(1, .3)
g race = int(runiform()*4+1)
lab def race 1 "Race1" 2 "Race2" ///
    3 "Race3" 4 "Race4", modify
lab val race race
lab def yn 0 "No" 1 "Yes", modify
lab val vocational unigrade yn
su *

table un vo ra

*use tab
    *combine categories, even if labeled
    g combined = .
    levelsof unigrade, loc(lu)
    levelsof vocation, loc(lv)
        loc labu `"`:value label unigrade'"'
        loc labv `"`:value label vocation'"'    
    loc ii = 1
    foreach u in `lu' {
    loc i =0
    foreach v in `lv' {
      loc labb `"`labb'  `ii' `"Uni:`:label `labu' `i''--Voc:`:label `labv' `i''"' "'
       replace combined = `ii' ///
            if unigrade ==`u' & vocation ==`v'
        loc `++i'
        loc `++ii'
    }
    }
        di in r `"`labb'"'
    lab def combined `labb', modify
    lab val combined combined
    
    ta  combined race, row

Produces a new variable 'combined' with these category labels (you can change these to make them more informative

.         lab list
combined:
           1 Uni:No--Voc:No
           2 Uni:Yes--Voc:Yes
           3 Uni:No--Voc:No
           4 Uni:Yes--Voc:Yes

This is how combined is related to (recoded from vocation and unigrade):

       ta combined unigrade

                 |       unigrade
        combined |        No        Yes |     Total
-----------------+----------------------+----------
  Uni:No--Voc:No |        44          0 |        44
Uni:Yes--Voc:Yes |        23          0 |        23
  Uni:No--Voc:No |         0         26 |        26
Uni:Yes--Voc:Yes |         0          7 |         7
-----------------+----------------------+----------
           Total |        67         33 |       100


.         ta combined vocation

                 |      vocational
        combined |        No        Yes |     Total
-----------------+----------------------+----------
  Uni:No--Voc:No |        44          0 |        44
Uni:Yes--Voc:Yes |         0         23 |        23
  Uni:No--Voc:No |        26          0 |        26
Uni:Yes--Voc:Yes |         0          7 |         7
-----------------+----------------------+----------
           Total |        70         30 |       100

Here's the final output:

.         ta  combined race, row nokey

                 |                    race
        combined |     Race1      Race2      Race3      Race4 |     Total
-----------------+--------------------------------------------+----------
  Uni:No--Voc:No |        12         11          7         14 |        44
                 |     27.27      25.00      15.91      31.82 |    100.00
-----------------+--------------------------------------------+----------
Uni:Yes--Voc:Yes |         7          7          6          3 |        23
                 |     30.43      30.43      26.09      13.04 |    100.00
-----------------+--------------------------------------------+----------
  Uni:No--Voc:No |         4          7          8          7 |        26
                 |     15.38      26.92      30.77      26.92 |    100.00
-----------------+--------------------------------------------+----------
Uni:Yes--Voc:Yes |         2          1          1          3 |         7
                 |     28.57      14.29      14.29      42.86 |    100.00
-----------------+--------------------------------------------+----------
           Total |        25         26         22         27 |       100
                 |     25.00      26.00      22.00      27.00 |    100.00

Note how this final table matches the frequencies and the row percentages produced in the earlier, separate -table- commands I provided. Now these freq/row percentages are all combined in this last table.