Standard deviation in panel data set

[Mridula Duggal]

Hello,

I have a Panel Dataset and I am looking to calculate standard deviations of a variable, that is, quarterly inflation, for 53 countries over a period of 1980-2014. Could you tell me how I might be able to perform the same? I need to calculate the standard deviation over two different periods. For example, between 1980 and 1998 and then from 1999 and 2014.

I have performed the following, however I am not sure how stata calculates it, as upon using it I get one value repeated value for SD for each country. I am using Stata 12 and 13 (both) .

 
 bys country : egen SD = sd(inflation)

Any help would be extremely useful,

Thank you,
Mridula

[Carlo Lazzaro]

Mridula:
what if you type:

bysort country year: egen SD=sd(inflation)

[Jorrit Gosens]

The way I read it is very close but a little different from what Carlo suggests. If you want to use only data in certain periods of years, you will have to tell Stata which are the periods you are interested in. First create a variable that indicates period, then do the calculation of sd simialr to what Carlo suggested, e.g.:

gen period = 1 if Year>=1980 & Year <=1998
replace period = 2 if Year>=1999 & Year <=2014
bysort country period: egen SD=sd(inflation)

[Mridula Duggal]

Carlo,

If I use the above code, it creates only missing value for standard deviations.

Thank you,
Mridula

[Mridula Duggal]

Jorrit,

I will try using what you have suggested.

Thank you,
Mridula

[Jorrit Gosens]

Do you have quarterly or yearly data? With quarterly data Carlo's code should at least also work.

[Mridula Duggal]

I have both quarterly and yearly data. However, at the moment working with yearly data.

[Jorrit Gosens]

Well, the code Carlo suggests would calculate sd per country and year. That's possible with quarterly data. With yearly data you're asking Stata to calculate the sd of a single value, producing a missing value.

[Mridula Duggal]

Right. Thank you so much. Will try this with quarterly data. Hopefully, should work.