displaying 14-digit numeric variable (double) in tab and codebook commands

[Rosa Blau]

Hello,

I have a 14-digit numeric variable, named 'year_id' (type double), which I would like STATA to display fully (ie. not in scientific notation), at least in the codebook. It's working with the table command, but not with the codebook or tab command (the latter of which I also use frequently). Importantly, I need this variable to remain numeric because I will be adding it to other id variables to creat unique ids for each of my 3million+ observations. Please see my code below:

Best of thanks for your consideration,
Rosa

[Nick Cox]

This is an example of the x-y problem. Unique identifiers for 3 million observations could just be

gen long id = _n 

Informative identifiers could easily be based on two or more variables using something like

egen long id = group(year whatever whoever somethingelse), label

[Rosa Blau]

Thank you! The egen code looks promising, but I began running it 30 minutes ago.. and it's still going...

I have Stata/IC with a 1.3 GHz Intel Core i5 processor, but still this seems a bit slow in a dataset with 'only' 3,156,487 observations and (including my new id variable) 24 variables. I'm starting to assume that something has gone wrong.

Is there no other way to have codebook display a 14-digit numeric variable, generated with a 'gen' command, in non-scientific notation?

[Maarten Buis]

Importantly, I need this variable to remain numeric because I will be adding it to other id variables to creat unique ids for each of my 3million+ observations.

You can concatenate ("add") strings to create a composite id variable. Since you are working with such large numbers and such large datasets, that is probably safer. Moreover, it will solve your problem with codebook. Here is an artificial example for a composite id for a panel dataset with 3 waves of three families with 4 persons per family.

// create example dataset
drop _all
set obs 3
gen wave = _n
expand 3
bys wave : gen fam_id = _n
expand 4
bys wave fam_id : gen pers_id = _n

// create the composite id
gen id = strofreal(wave) + "_" + strofreal(fam_id) + "_" + strofreal(pers_id)

// admire the result
list, sepby(wave fam_id)

You don't have to store the parts that make up the final id (important in large datasets), as you can recover them easily with the split command, in this case split id, gen(id) parse(_) would do the trick.

[Rosa Blau]

I see! So I could convert the numeric variables to strings and then concatenate the strings. Thank you, I will try that.

[Nick Cox]

With a machine that I guess is fairly humdrum, I was getting results in seconds:

. clear

. set obs 3000000
number of observations (_N) was 0, now 3,000,000

. gen year = runiformint(1991, 2018)

. gen whatever = cond(_n < 1.5e6, "A", "B")

. set rmsg on
r; t=0.00 10:40:56

. egen id = group(whatever year), label
r; t=6.37 10:41:18

Most of the time there was me thinking and wanting coffee. Your calculation may well be more complicated, but it shouldn't take that long.

In my case there were 56 distinct classes, all 28 possible years by 2 categories. You may well have many, many more, but I am still surprised at 30 minutes +.

[Rosa Blau]

I too am surprised. Out of impatience I'm using my unique id without the help of codebook for the moment, but I will try the egen command again later.

[Rosa Blau]

Dear Statalisters,

I am once again having a similar problem, although it is simpler this time. I am using a 10-digit numeric (double) variable. I am not permitted to give a data example, but the values are all integers. I have tried the formats %10.0g and %10.0f, but the variable continues to display in scientific notation. I have studied the manual entry on format, but clearly I am missing something. I greatly appreciate your time and advice on this issue.

Best wishes,
Rosa

[Nick Cox]

You are going to have to make up an example that doesn't work. Can you reproduce this?

. clear

. set obs 1
number of observations (_N) was 0, now 1

. gen double y = 9876543210

. l

     +-----------+
     |         y |
     |-----------|
  1. | 9.877e+09 |
     +-----------+

. format y %12.0f

. l

     +------------+
     |          y |
     |------------|
  1. | 9876543210 |
     +------------+

. format y %11.0f

. l

     +------------+
     |          y |
     |------------|
  1. | 9876543210 |
     +------------+

. format y %10.0f

. l

     +------------+
     |          y |
     |------------|
  1. | 9876543210 |
     +------------+

[Maarten Buis]

In the display format one character is reserved for the decimal point and one for the sign, so if you want to display 10 digits you need to use the format %12.0g or %12.0f:

. di %10.0g 1234567890
 1.235e+09

. di %12.0g 1234567890
  1234567890

[Rosa Blau]

Dear Maarten, dear Nick,

With all formats listed in your posts I can get the full 10-digit integer display with the command 'table' (but not with tab or codebook). I think I should be able to work with this for now, so best of thanks for your advice.

Best wishes, Rosa

[Nick Cox]

OK, but there was no mention of tabulate or codebook in #8. I don't know an general work-around for codebook but I have got to suggest that not seeing sample identifiers exactly wouldn't worry me for what I use codebook for.

For tabulate there could be many alternatives or work-arounds, but all depends on what you are tabulating, and we need to see details.

That said, with 10-digit identifiers I experience no difficulty if use a string version:

. clear

. set obs 1
number of observations (_N) was 0, now 1

. gen double nid = 9876543210

. gen sid = string(nid, "%10.0f")

. codebook

--------------------------------------------------------------------------------------------------------------------------------------------------
nid                                                                                                                                    (unlabeled)
--------------------------------------------------------------------------------------------------------------------------------------------------

                  type:  numeric (double)

                 range:  [9.877e+09,9.877e+09]        units:  1000
         unique values:  1                        missing .:  0/1

            tabulation:  Freq.  Value
                             1  9.877e+09

--------------------------------------------------------------------------------------------------------------------------------------------------
sid                                                                                                                                    (unlabeled)
--------------------------------------------------------------------------------------------------------------------------------------------------

                  type:  string (str10)

         unique values:  1                        missing "":  0/1

            tabulation:  Freq.  Value
                             1  "9876543210"

. gen y = 42

. tab sid, su(y)

            |            Summary of y
        sid |        Mean   Std. Dev.       Freq.
------------+------------------------------------
 9876543210 |          42           0           1
------------+------------------------------------
      Total |          42           0           1

.