tabulation with too many values

[Charles Bondi]

I have 60 categories for which I want to portray values of a discrete variable taking 8 values in either tabular form. So tab and tab2 respond with "too many values". It's important for me to identify categories for which I have zero quantity of my variable. With table commands apparently failing because of the number of categories, any ideas about how to proceed?

[Clyde Schechter]

Although tabulate has a limit on the number of values it will accept, -table- does not:

table category variable_with_8_values

will run without complaint.

[Sergiy Radyakin]

Clyde's advice is good, table is a much more versatile command, but 60 categories is within the reach of tabulate as well:

sysuse auto, clear
generate x=_n
expand 8

generate y= floor(_n/74)
tabulate x
tabulate x y

So, Charles, may be check your data if your expectations about the number of categories is correct. Inspect and levelsof could be the tools for this task.

Best, Sergiy Radyakin

[Nick Cox]

groups (Stata Journal) has a handle specifically for showing cross-combinations with zero frequency. See https://www.statalist.org/forums/for...updated-on-ssc for some details.

[Charles Bondi]

Thanks to all of you. I'll try these when I'm back on site with the data.

[David Radwin]

You might also try bigtab by Paul H. Bern, available from SSC.

. ssc describe bigtab

---------------------------------------------------------------------------------------------------------
package bigtab from http://fmwww.bc.edu/repec/bocode/b
---------------------------------------------------------------------------------------------------------

TITLE
      'BIGTAB': module to produce frequency tables for "too many values"

DESCRIPTION/AUTHOR(S)
      
      The bigtab command helps when you get a "too many values" error
      from tabulate or want to save the results in a data file.  Users
      may specify up to   three variables and have the option of
      displaying and saving cumulative, row, and column   counts and
      percentages.
      
      Distribution-Date: 20030609
      
      Author: Paul H. Bern, Princeton University
      Support: email [email protected]
      

INSTALLATION FILES                               (type net install bigtab)
      bigtab.ado
      bigtab.hlp
---------------------------------------------------------------------------------------------------------
(type ssc install bigtab to install)

[Sergiy Radyakin]

Another approach is to collapse the dataset basically to the dimensions of the table:

sysuse nlsw88, clear

collapse (count)freq = idcode, by(wage)
gsort -freq wage

list

Results in a dataset of frequencies:

. list

     +-----------------+
     |     wage   freq |
     |-----------------|
  1. | 4.025765     48 |
  2. | 5.032206     36 |
  3. |  8.05153     27 |
  4. | 4.830918     25 |
  5. | 5.636071     24 |
     |-----------------|
  6. | 3.220612     23 |
  7. | 3.344482     22 |
  8. | 4.180602     19 |
  9. | 10.06441     19 |
 10. | 4.227053     18 |
     |-----------------|
 11. | 7.745568     18 |
 12. | 3.623188     17 |
....cut here..................
959. | 29.72623      1 |
960. | 30.19324      1 |
     |-----------------|
961. | 30.33817      1 |
962. | 30.92161      1 |
963. | 30.96618      1 |
964. | 30.96741      1 |
965. |  33.4984      1 |
     |-----------------|
966. | 39.23074      1 |
967. | 40.74659      1 |
     +-----------------+

Having 48 persons out of 2,246 with same wages up to millionths fractions of a dollar (and not a round number) in a real dataset is, imho, phenomenal. If anyone knows how this has happened, please share.