Means

[Joe Ward]

Dear All

I have data on GDP (var= gdp) for different countries (var = iso) , for different years (var=year). My data is in long view. I would like to generate a new variable that calculates the average GDP over the proceeding 10 years for each year for each country. I know i need egen and by sort, but am struggling! Any help would be much appreciated!

Many Thanks

Joe

[Nick Cox]

Many threads can be found here on such problems with a little searching.

egen could be used, but I think you'd need to install user-written functions to make easy headway.

A more general approach would be based on installing tsegen (SSC). Notional example:

webuse grunfeld

xtset company year

tsegen invest10 = rowmean(L(1/10).invest)

edit company year invest*

That calculation uses whatever data are available, even single values. See also the help.

See original thread at http://www.statalist.org/forums/foru...egen-functions and if needed search for further mentions.

[Joe Ward]

Hi Nick
Thank you so much for your quick reply. Looking at your example it seems this is what I need, but it still doesn't work with my data, as there are repeated time values within the panel; each country has mortality data for each year, for each age group within each year and finally for each sex within each age group. For each mortality observation, I need the corresponding mean gdp from the previous 10 years - I hope that makes sense. Here is an example:

country year ****age **sex ****mortality** gdp
a *******1990**** 1**** m******** 6 *******10
a *******1990**** 1**** f *********5 *******10
a *******1990**** 2 ****m******** 2******* 10
a *******1990**** 2**** f **********4 *******10
a *******1991**** 1**** m********* 8******* 12
a *******1991**** 1**** f **********7******* 12
a *******1991**** 2**** m *********3******* 12
a *******1991**** 2**** f **********4******* 12

(sorry about the stars, its the only way I could draw the table!)

Any help would be much appreciated!

Best Wishes

Joe

[Richard Williams]

That sounds much more complicated than what you originally stated. My guess is that, instead of using country as the ID variable, you need to create a new id variable for each unique combination of country/age/sex/ etc. Or something like that.

[Joe Ward]

Hello
thank you again for your help - Ive worked out a way of doing it. I had merged the gdp data (which just has one observation per year per country) with the mortality data. I just used your code to get my 10 year gdp average on the original data set, then merged it with the mortality data - I think it worked!
Thank you