Help with the creation of a lagged standard deviation variable

[Daniel Davis]

Hello all,

I am hoping for a bit of help with the generation of a new variable. I have some experience with STATA, but I am still very new to the software.

Although my data contains many variables for this question only three variables are important. Firm, Year, and Score. I am wanting to create a new variable that is the Standard Deviation of the Score variable that lagged by one year and includes the past 3 years scores to compute the standard deviation. This calculation needs to be by firm. I know I may not be communicating this clearly, so I am including some sample data to illustrate my goal.

Firm	Year	Score
A	1991	90
A	1991	95
A	1992	96
A	1992	92
A	1993	93
A	1993	94
A	1994	96
A	1994	93
A	1995	95
A	1995	94
A	1996	94
A	1996	87
A	1996	93
A	1997	94
A	1998	90
A	1999	90
A	1999	97
A	1999	91
A	2000	92
A	2001	92
A	2001	87
A	2002	94
B	1995	89

This is a small sample of my data and includes the first firm in the data set. Let's call the new variable SDScore. For the row A 2002 94, the value of SDScore should be the standard deviation of the values of Score in the Year 2001 2000 and 1999. Similar logic for the other years until the year 1993. For the row A 1993 94 and the row A 1993 93 the value of SDScore should be the standard deviation of the Score values in the Year 1992 and 1991. Finally for 1992 and for 1991 SDScore should be calculated using only the 1991 values for Score.

Finally, it is important to note that not every firm's observations are in the same year. For instance, firm B has observations that go back to 1963.

Are there any commands I can use to calculate SDScore in my dataset? Thanks!

Best,
Daniel

[Nick Cox]

Various programs will do this for you. That most familiar to me is rangestat from SSC. You must install it with

ssc install rangestat

rangestat (sd) wanted=score, int(year -3 -1) by(id)

is a guess at syntax.

FWIW, 3 observations seems rather small to get a good idea of SD.