egen for specific time period

[Souryabrata Mohapatra]

Hi,

I want to calculate the standard deviation for a specific year group i.e.-

year	variable	SD_variable
1990	1	a
1991	2	a
1992	3	a
1993	4	b
1994	5	b
1995	6	b

There's countrycode also which implies I need to use egen

> bysort countrycode: egen SD_variable= sd(variable)

But the above command calculate SD from 1990-95

Where I want to calculate the same for a specific periods- 1990-92 & 1993-95


Can you please help?

[Souryabrata Mohapatra]

Is it correct-
egen meangdp = mean(gdp) if inrange(year,1963,1980), by(country)

[Joro Kolev]

From what you show, the code is

egen  meangdp = mean(gdp), by(SD_variable)

but it is also possible that you re not explaining very well what you want. In particular do you have this SD_variable, or you want to calculate it?

[Nick Cox]

That could be correct but over any long period a geometric mean might make more sense. For periods 1990-1992, 1993-1995 consider this

. clear

. set obs 6
number of observations (_N) was 0, now 6

. gen whatever = _n

. gen year = 1989 + _n

. egen mybin = seq(), block(3)

. egen mean = mean(whatever), by(mybin)

. l, sep(3)

     +--------------------------------+
     | whatever   year   mybin   mean |
     |--------------------------------|
  1. |        1   1990       1      2 |
  2. |        2   1991       1      2 |
  3. |        3   1992       1      2 |
     |--------------------------------|
  4. |        4   1993       2      5 |
  5. |        5   1994       2      5 |
  6. |        6   1995       2      5 |
     +--------------------------------+

and you naturally may work with blocks for periods and countries at once, say

egen mean2 = mean(whatever), by(country mybin)

Much more at

SJ-18-3 dm0095 . . . . . . . . . . . Speaking Stata: From rounding to binning
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
Q3/18 SJ 18(3):741--754 (no commands)
basic review of how to bin variables in Stata, meaning how to
divide their range or support into disjoint intervals

https://www.stata-journal.com/articl...article=dm0095

[Souryabrata Mohapatra]

Hi Joro Kolev,
I want to calculate SD_variable
I think inrange could be a good way to go
So the command can be:

egen meangdp = mean(gdp) if inrange(year,1963,1980), by(country)

?

[Joro Kolev]

I think you want to do what Nick was explaining above.

You firstly need to create a variable that is containing your bins for the years. For example in the initial data that you showed, what you called SD_variable was containing one value (a) for the first three years, and another value (b) for the other three years. You need to create such a variable, because you know what are your bins, but I do not know them, neither does Stata know them.

There are more clever ways how to do this, but to tell you the clever ways, you need to say exactly what your years are and how you want to bin them.

Brute force is always an option:

gen bin = 1 if inrange(year,1963,1980)
replace bin = 2 if inrange(year,1981,1985)
replace bin = 3 if inrange(year,1986,1990)
...
and so on until you have created all the bins

Once you have the bins

egen meangdp = mean(gdp), by(country bin)

will give you the mean GDP in the country within the respective bin, and

egen sdgdp = sd(gdp), by(country bin)

will give you the standard deviation of GDP in the country within the respective bin.

Hi Joro Kolev,
I want to calculate SD_variable
I think inrange could be a good way to go
So the command can be:

egen meangdp = mean(gdp) if inrange(year,1963,1980), by(country)

?

[Souryabrata Mohapatra]

Hi Nick Cox,
What I did is

egen SD_variable1 = sd(variable) if inrange(year,1990,1992), by(id)
egen SD_variable2 = sd(variable) if inrange(year,1993,1995), by(id)
gen SD_variable = max(SD_variable1, SD_variable2)
drop SD_variable1 SD_variable2

Is that okay?

[Nick Cox]

So your SDs are based on most 3 values?

You could do this

gen period = cond(inrange(year, 1990, 1992), 1 , cond(inrange(year, 1993, 1995), 2, .)) 
egen SD = SD(variable), by(id period)

[Souryabrata Mohapatra]

No Nick Cox,
That was just an example to ask & clarify my doubts.
I've 30 years of panel data.
I divided it into two periods, so 15 years each.