error for displaying a large number

[Joonho Kim]

Hello,

To merge, I need to make a specific identifier for each row as you can see below

clear
ccode1 ccode2 year dyad
2 20 1920 2020
2 20 1921 2020
2 20 1922 2020
2 20 1923 2020
2 20 1924 2020
end

I executed the code like this

"gen dyad_year=ccode1*1000000000 + ccode2*10000 + year"

What I expected to see in dyad_year for the first row is

2000201920

and the second row in dyad_year is

2000201921

and same for the remaining rows.


But, when I executed the code above, the results like this

ccode1 ccode2 year dyad dyad_year
2 20 1920 2020 2000201984
2 20 1921 2020 2000201984
2 20 1922 2020 2000201984
2 20 1923 2020 2000201984
2 20 1924 2020 2000201984
end

I am wondering why my code does not work.

Thank you

[Clyde Schechter]

Because your -gen- command did not specify a storage type, Stata automatically creates the new variable as a float. But floats do not have enough bits to accurately represent 10 digit numbers, so you are experiencing precision errors. See -help precision- for more information. The solution is to force Stata to use a data type with greater precision:

gen long dyad_year=ccode1*1000000000 + ccode2*10000 + year

Also, depending on how you plan to use this identifier, it might be better to make it a string variable using the -egen, concat()- command. See -help egen-. (But don't do this if you will need to do any calculations with it, or if you will need to use it in an -xtset- or -tsset- command.)

[Sarah Edgington]

It's because you're creating a variable stored as float by default and the largest integer a float can hold is 16,777,216. See help data_types . You need to create the variable as double.

You can get the result you expect from:

gen double dyad_year=ccode1*1000000000 + ccode2*10000 + year
format dyad_year %11.0g

That said, there's no reason to create this variable for a merge. Just use multiple variables in your merge command.

merge 1:1 ccode1 ccode2 year using "filename"

[Joonho Kim]

Thank you. It works perfectly.