How to writing a code similar as the following code but by region?

[Linghui Han]

levelsof year, local(years)
egen flag = tag(id_job year)
gen stateratio = .
foreach y of local years {
summ occ if year == `y' & flag, meanonly
local odds = `r(mean)'*100
replace stateratio = `odds' if year == `y'
}

[Linghui Han]

Or, I need to get the value for the variable stateratio by region by year. What will be the efficient code? Thank you!

[Nick Cox]

The original (no data example!) appears equivalent to

egen flag = tag(id_job year)
rangestat stateratio = occ if flag, int(year 0 0)
bysort year (stateratio) : replace stateratio = 100 * stateratio[1]

where rangestat is from SSC. That should make it easier for you.

Alternatively, this should be quicker than what you have, but slower than rangestat:

egen flag = tag(id_job year)
egen stateratio = mean(cond(flag, 100 * occ, .)), by(year)

noting that 100 times the mean of something is the mean of 100 times something).

[Linghui Han]

Thank you very much! I changed your code to the following:

egen flag = tag(id_job year)
rangestat stateratio = occ if flag, int(year 0 0) by(province)
bysort year (stateratio) : replace stateratio = 100 * (stateratio[1]/100)

[Nick Cox]

I don't see much point to multiplying by 100/100.

[Linghui Han]

What does [1] in stateratio[1] mean?

[Linghui Han]

Without dividing by 100, the number I got equal 4000 if the percentage point Is 40.

[Linghui Han]

But from "rangestat stateratio = occ if flag, int(year 0 0) by(province)", the number is 0.4

[Nick Cox]

But 100/100 is 1. Multiplying by 1 is harmless but pointless. As for [1], it is a subscript and identifies the first observation in each group.

Note that we can talk about code. We can't talk about your data and results if we can't see them.

[Linghui Han]

I change code to:

rangestat state = occ if flag, int(year 0 0) by(province)
bysort province year (state) : replace state = state*100

With state[1] the data change the value of the variable "state" in the first observation (e.g. 0.05) by multiplying 100 (=100*0.05) and the value in the second observation by multiplying 100 and the changed value of the first observation (=100*5).