codebook differences between %td and %tc formats

[Melissa Saphir]

Dear Stata community:

I am wondering whether there is a bug in the Stata 17 codebook command affecting what it indicates about the %tc date-time format, or whether I am misunderstanding date-time formats?

Background: When I have applied the %td format to a numeric variable, the codebook tells me that the variable is a daily date, as shown with date2 in the example below. This makes sense to me--I have a numeric variable, and because I applied the %td format, Stata knows it is a daily date, and the codebook tells me so.

. codebook date2 // Starting with a numeric variable.

------------------------------------------------------------------------------------------------------------
date2 (unlabeled)
------------------------------------------------------------------------------------------------------------

Type: Numeric (float)

Range: [22471,22986] Units: 1
Unique values: 49 Missing .: 0/76

Mean: 22900.8
Std. dev.: 103.416

Percentiles: 10% 25% 50% 75% 90%
22883 22894 22922.5 22953.5 22978

. format date2 %td // Applying daily date formatting so the variable is human-readable.

. codebook date2 // Output is as I expect--Yay!

------------------------------------------------------------------------------------------------------------
date2 (unlabeled)
------------------------------------------------------------------------------------------------------------

Type: Numeric daily date (float)

Range: [22471,22986] Units: 1
Or equivalently: [10jul2021,07dec2022] Units: days
Unique values: 49 Missing .: 0/76

Mean: 22900.8 = 12sep2022(+ 18 hours)
Std. dev.: 103.416
Percentiles: 10% 25% 50% 75% 90%
22883 22894 22922.5 22953.5 22978
26aug2022 06sep2022 04oct2022 04nov2022 29nov2022


In contrast, when I apply %tc formatting to a numeric variable, the codebook does NOT indicate it is a date-time variable, even though the variable displays as a date-time the way I want it to when using the -browse- or -list- commands, as shown with date1 below. Why doesn't the -codebook- command produce output with a human-readable date1?

. codebook date1 // Starting with a numeric variable.

------------------------------------------------------------------------------------------------------------
date1 (unlabeled)
------------------------------------------------------------------------------------------------------------

Type: Numeric (float)

Range: [1.977e+12,1.987e+12] Units: 100000
Unique values: 65 Missing .: 10/76

Mean: 2.0e+12
Std. dev.: 3.0e+09

Percentiles: 10% 25% 50% 75% 90%
2.0e+12 2.0e+12 2.0e+12 2.0e+12 2.0e+12

. format date1 %tc // Applying date-time formatting so the variable is human-readable.

. codebook date1 // Output is not as I expect, there is no indication that date1 is formatted as a date-time variable.

------------------------------------------------------------------------------------------------------------
date1 (unlabeled)
------------------------------------------------------------------------------------------------------------

Type: Numeric (float)

Range: [1.977e+12,1.987e+12] Units: 100000
Unique values: 65 Missing .: 10/76

Mean: 2.0e+12
Std. dev.: 3.0e+09

Percentiles: 10% 25% 50% 75% 90%
2.0e+12 2.0e+12 2.0e+12 2.0e+12 2.0e+12


. list date1 // However this output IS as I expect! Yay! (But why doesn't the codebook reflect the %tc formatting?)

+--------------------+
| date1 |
|--------------------|
1. | 29aug2022 10:28:25 |
2. | 30aug2022 09:57:27 |
.
.
.

I would appreciate any insight into the differences in how the codebook displays %tc vs. %td variables!
Melissa

[William Lisowski]

I agree that codebook does not indicate that your date1 variable is a clock date variable in the same way that it indicates date2 is a daily date variable.

I initially thought it might be because your date1 variable is incorrectly stored as a float, which is not appropriate for a clock date variable - help datetime tells us clock values must be stored in double variables. But the following example shows us that even stored as a double, codebook does not identify it as a time variable.

* Example generated by -dataex-. For more info, type help dataex
clear
input float var2
22471
22986
end

gen date2 = var2
format date2 %td

generate var1 = dhms(date2,8,20,30) // incorrect
generate date1 = var1
format date1 %tc

generate double date1d = dhms(date2,8,20,30)
format %tc date1d

list
describe
codebook

. list

     +------------------------------------------------------------------------+
     |  var2       date2       var1                date1               date1d |
     |------------------------------------------------------------------------|
  1. | 22471   10jul2021   1.94e+12   10jul2021 08:20:47   10jul2021 08:20:30 |
  2. | 22986   07dec2022   1.99e+12   07dec2022 08:19:36   07dec2022 08:20:30 |
     +------------------------------------------------------------------------+

. describe

Contains data
 Observations:             2                  
    Variables:             5                  
------------------------------------------------------------------------------------------------
Variable      Storage   Display    Value
    name         type    format    label      Variable label
------------------------------------------------------------------------------------------------
var2            float   %9.0g                 
date2           float   %td                   
var1            float   %9.0g                 
date1           float   %tc                   
date1d          double  %tc                   
------------------------------------------------------------------------------------------------
Sorted by: 
     Note: Dataset has changed since last saved.

. codebook

------------------------------------------------------------------------------------------------
var2                                                                                 (unlabeled)
------------------------------------------------------------------------------------------------

                  Type: Numeric (float)

                 Range: [22471,22986]                 Units: 1
         Unique values: 2                         Missing .: 0/2

            Tabulation: Freq.  Value
                            1  22471
                            1  22986

------------------------------------------------------------------------------------------------
date2                                                                                (unlabeled)
------------------------------------------------------------------------------------------------

                  Type: Numeric daily date (float)

                 Range: [22471,22986]                 Units: 1
       Or equivalently: [10jul2021,07dec2022]         Units: days
         Unique values: 2                         Missing .: 0/2

            Tabulation: Freq.  Value
                            1  22471  10jul2021
                            1  22986  07dec2022

------------------------------------------------------------------------------------------------
var1                                                                                 (unlabeled)
------------------------------------------------------------------------------------------------

                  Type: Numeric (float)

                 Range: [1.942e+12,1.986e+12]         Units: 100000
         Unique values: 2                         Missing .: 0/2

            Tabulation: Freq.  Value
                            1  1.942e+12
                            1  1.986e+12

------------------------------------------------------------------------------------------------
date1                                                                                (unlabeled)
------------------------------------------------------------------------------------------------

                  Type: Numeric (float)

                 Range: [1.942e+12,1.986e+12]         Units: 100000
         Unique values: 2                         Missing .: 0/2

            Tabulation: Freq.  Value
                            1  1.942e+12
                            1  1.986e+12

------------------------------------------------------------------------------------------------
date1d                                                                               (unlabeled)
------------------------------------------------------------------------------------------------

                  Type: Numeric (double)

                 Range: [1.942e+12,1.986e+12]         Units: 100000
         Unique values: 2                         Missing .: 0/2

            Tabulation: Freq.  Value
                            1  1.942e+12
                            1  1.986e+12

.