Problem manipulating large database based on conditional statements to eliminate specific observations

[Kiethiyoth Khaou]

Hello world of stats and analysis,

I have dibbled a bit in STATA over my university years but my work has required me to become much more engaged with the software.

I am currently working on a project (using STATA 13.0) trying to determine seasonal sales patterns to implement a sales sampling strategy across multiple provinces in Cambodia.

I have a large dataset with 24 variables and 120,000+ observations. My first step is to manipulate the current dataset to remake sales years based on seasonal patterns in Cambodia. It happens that a sales year makes much more sense from November 1st - October 31st (the following year). The dataset has sales records of every sale done by any supplier based on date (SUPPLIER_ID / DAY / MONTH/ YEAR / SALE_AMOUNT).

I am having problems writing code that would systematically eliminate any sales before November the year that particular supplier joined our records. For example supplier 43 joined in 01/04/2014 and has sales until 01/09/2017. I would like to delete all sales previous to 01/11/2014 for supplier 43, and then do that for all suppliers in my dataset.

I have been fiddling with the bysort function: bysort SPID year month: generate y=1 then replace y=. if SPID==SPID[_n-1] & month>=11. I feel like im close but just missing the code to tell STATA to identify and replace only the first year they joined so that later I can write a code to eliminate all 'missing values'

I hope all that makes sense,

Thank you all for your help!

[Clyde Schechter]

by SPID (year month), sort: drop if year == year[1] & month < 11

will eliminate those observations altogether. To just mark them with a variable:

by SPID (year month), sort: gen byte mark = (year == year[1] & month < 11)

[Kiethiyoth Khaou]

That worked swimmingly. Thank you so very much Clyde!