Re: st: D. Time series operator

[Michael Hanson <mshanson@mac.com>]

On Nov 13, 2008, at 1:16 PM, Diana Eastman wrote:

> I know that D., generates differences of any order. Could someone
> explain how to interpret
> D(0/1).(dummyvar1 dummyvar2)
>
> Specifically, what does (0/1) do with regards to the differencing?
>
> The context of this is an instrumental variable regression.


Diana:

What you are looking for -- admittedly somewhat tricky to find -- is - 
help tsvarlist- to explain this syntax.  See also <http:// 
www.stata.com/support/faqs/stat/lags.html>.  (You can find both of  
these resources by typing -findit time series operators- in Stata,  
although it does take a little digging through the resulting list of  
items.)

D(0/1).(x y) expands to D(0/1).x D(0/1).y, which in turn expands to  
D0.x D1.x D0.y D1.y -- that is, the "zeroth" and first differences of  
the time series variables x and y.  (And although you mentioned you  
understand the D. operator, for the sake of posterity I should note  
that the "zeroth" difference of a series is just that series: D0.x is  
simply x.  Also, D1.x and D.x are synonymous in Stata.)

Another example of this syntax:  L(1/4).z expands to L.z L2.z L3.z  
L4.z .  All told, D(0/1).(x y) is just a convenient shorthand for  
specifying several time series operations on several variables all at  
once.  This syntax saves time and typing with VARs, ADL models, etc.  
-- as well as long lists of lagged instruments for IV or GMM estimation.

More generally, the "/" between two numbers is used in Stata notation  
to denote a sequence of integers.  For example, consider the syntax:

	forvalues i = 1/4 {
		generate L`i'_z = L`i'.z
	}

Here, `i' takes on the values 1, 2, 3, 4 in turn.  This code snippet  
creates four new time series variables: L1_z through L4_z, which  
equal the first through fourth lags of the variable z.  (One might  
want to do this to use lagged series with Stata commands that do not  
accept the lag operator notation.)


Hope that helps,
Mike

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/