regression results changes every time I ran it

[Olivia Li]

Hi, everyone,
I feel confused that each time I ran my code to generate the data and also regreesion, the regeression results changes a little bit. Not much, but always change.
I didn't generate random numbers.

Thanks so much if any advice could be offered. Thanks!

[Rich Goldstein]

are you -sort-ing the data? if there are ties, sort will break the tie haphazardly (some might say randomly) unless you use the "stable" option; see

h sort

if you are not sorting the data then you need to show the code (at least for me - maybe someone else has another guess)

[Olivia Li]

are you -sort-ing the data? if there are ties, sort will break the tie haphazardly (some might say randomly) unless you use the "stable" option; see

h sort

if you are not sorting the data then you need to show the code (at least for me - maybe someone else has another guess)

Thanks so much, Rich!
However, concerning "sort" code, I just used "bys city year: egen X_median = median(X) " and "bys city year: egen X_mean = mean(X) " and "bys city year: egen X_total = total(X) ".
Would this cause any problems?

[Jared Greathouse]

Show me the regression code. There's no reason you should be getting different results each time unless you're generating random numbers and not using the same seed each time

[Hemanshu Kumar]

Olivia Li as requested in #2 and #4, we need to see your code, and preferably a data extract (for this, use -dataex-) and be able to replicate your issue, to figure out what's going on.

In addition, note that you can avoid sorting your data by using -by()- as an option to your egen command, rather than using it as a prefix.

[Carlo Lazzaro]

Olivia:
as an aside to previous excellent advice: do -summarize- results also change each time you run your code?