Values of variable changed after saving and reopening .dta file

[Francesco Zaccardi]

Dear Stata Users,

I am encountering an issue I have not been able to explain and would appreciate your insights (details/example below).

I am working with a large dataset (~20 million observations) that includes a variable d10, of type byte (%8.0g), which contains some missing values. After saving and reopening the dataset, there are new values (and a different distribution of d10).

Has anyone experienced something similar or can suggest a possible explanation?

Many thanks in advance,
Francesco

PS: If repeated, the distribution changes again.


~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
tab d10, m /*only frequencies*/

d10 | Freq.
------------+-----------------------------------
1 | 1,831,686
2 | 1,769,036
3 | 1,934,389
4 | 1,914,070
5 | 1,894,011
6 | 2,044,012
7 | 2,180,404
8 | 2,192,355
9 | 2,047,698
10 | 1,734,772
. | 37,855
------------+-----------------------------------
Total | 19,580,288

save "dataset0_test.dta", replace


use "dataset0_test.dta", clear
tab d10, m

d10 | Freq.
------------+-----------------------------------
0 | 409,325
1 | 1,793,906
2 | 1,733,184
3 | 1,892,326
4 | 1,874,283
5 | 1,853,590
6 | 2,002,021
7 | 2,137,323
8 | 2,148,649
9 | 2,002,380
10 | 1,696,193
. | 37,108
------------+-----------------------------------
Total | 19,580,288

[Hua Peng (StataCorp)]

Can you try:

use "dataset0_test.dta", clear
cf d10 using dataset0_test.dta
save "dataset0_test.dta", replace
use "dataset0_test.dta", clear
cf d10 using dataset0_test.dta

[Francesco Zaccardi]

Thanks for your suggestion, and apologies for the delayed response.

This issue has also affected a few colleagues recently, but the good news is that we are no longer able to reproduce it. It appears the problem was linked to recent changes at the university IT system level, specifically during the data migration from the old to the new infrastructure.

Now that the migration is complete, the issue seems to have resolved itself.

Thanks again,
Francesco