Stata doesn't save properly?

[Miguel Vazquez]

Unfortunately I cannot provide a MWE due to the confidentiality and size of the data. But here is a copy-paste of my console of what is happening. Essentially I'm appending several files together and everything is fine. Then I save the combined files into a new file and upon reloading the same file some entries have simply been changed. I can't make sense of this. Can somebody explain?

. do "/tmp/SD162064.000000"

.     * Combine years (2008-2017, 2016 not given)
.     use              "$pathEst/input/cleaning_intermediate/F101_2008_data.dta", clear

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2009_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2010_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2011_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2012_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2013_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2014_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2015_data.dta"

.     qui append using "$pathEst/input/cleaning_intermediate/F101_2017_data.dta"

.    
.     * Store intermediate file
.     save $pathEst/input/cleaning_intermediate/F101_data_all_years.dta, replace
file
    ~/data/transactions_ecuador/3_mivazq/Masters_Thesis/estimation/input/cleaning_intermedi
    > ate/F101_data_all_years.dta saved

.
end of do-file

. tab year

ANIO_FISCAL |      Freq.     Percent        Cum.
------------+-----------------------------------
       2008 |    122,408       11.20       11.20
       2009 |    128,116       11.73       22.93
       2010 |    122,497       11.21       34.15
       2011 |    105,965        9.70       43.85
       2012 |    118,403       10.84       54.68
       2013 |    121,721       11.14       65.83
       2014 |    128,361       11.75       77.58
       2015 |    117,485       10.75       88.33
       2017 |    127,488       11.67      100.00
------------+-----------------------------------
      Total |  1,092,444      100.00

. clear all

. use "$pathEst/input/cleaning_intermediate/F101_data_all_years.dta", clear

. tab year

ANIO_FISCAL |      Freq.     Percent        Cum.
------------+-----------------------------------
          0 |        970        0.09        0.09
       2008 |    122,408       11.20       11.29
       2009 |    128,116       11.73       23.02
       2010 |    122,497       11.21       34.23
       2011 |    105,430        9.65       43.89
       2012 |    118,403       10.84       54.72
       2013 |    121,721       11.14       65.87
       2014 |    128,361       11.75       77.62
       2015 |    117,485       10.75       88.37
       2017 |    127,053       11.63      100.00
------------+-----------------------------------
      Total |  1,092,444      100.00

.

Btw. not only "year" variable has been set to zero. ALL variables are set to zero for those 970 entries.


Interestingly, if I perform

cf _all using "$pathEst/input/cleaning_intermediate/F101_data_all_years.dta"

after opening the file, then it doesn't return anything (implying the dataset loaded is identical to the one stored). But this is funny because I tabulated "year" AFTER saving it, and there were no zero entries...


Thanks a lot!

EDIT: I'm using Stata 18.0 MP—Parallel Edition (4 core)

[Andrew Musau]

If you have a legal copy of Stata, contact Technical Services as Stata is not supposed to change observations in saved datasets. See https://www.stata.com/support/tech-support/contact/ on how to do so.