Stata 15 help for tsvarlist


[U] 11.4.4 Time-series varlists


Time-series varlists are a variation on varlists of existing variables. When a command allows a time-series varlist, you may include time-series operators. For instance, L.gnp refers to the lagged value of variable gnp. The time-series operators are

Operator Meaning --------------------------------------------------------- L. lag (x_t-1) L2. 2-period lag (x_t-2) ... F. lead (x_t+1) F2. 2-period lead (x_t+2) ... D. difference (x_t - x_t-1) D2. difference of difference (x_t - 2x_t-1 + x_t-2) ... S. "seasonal" difference (x_t - x_t-1) S2. lag-2 (seasonal) difference (x_t - x_t-2) ... ---------------------------------------------------------


1. Time-series operators may be repeated and combined. L3.gnp refers to the third lag of variable gnp, as do LLL.gnp, LL2.gnp, and L2L.gnp.

2. The lead operator F. is the complement of the lag operator L., so FL.gnp is just gnp.

3. Note that D1.gnp equals S1.gnp, but D2.gnp does not equal S2.gnp because

D2.gnp = (gnp_t - gnp_t-1) - (gnp_t-1 - gnp_t-2) = gnp_t - 2*gnp_t-1 + gnp_t-2


S2.gnp = (gnp_t - gnp_t-2)

4. Operators may be typed in lowercase or uppercase.

5. Operators can be typed in any order; Stata will convert the expression to its canonical form. For example, Stata will convert ld2ls12d.gnp to L2D3S12.gnp.

6. In addition to operator#, Stata understands operator(numlist) to mean a set of operated variables. For example, specifying


in a varlist is the same as typing

L.gnp L2.gnp L3.gnp

7. In operator#, making # zero returns the variable itself. Thus, instead of typing

regress y x L(1/3).x

you can save a few keystrokes by typing

regress y L(0/3).x

8. Before using time-series operators, you must declare the time variable using tsset.

. list l.gnp time variable not set r(111) . tsset time (output omitted) . list l.gnp (output omitted)

9. Variable names can be abbreviated subject to the usual Stata conventions. If the only variable in your dataset that begins with gn is gnp, then you can type instead of L.gnp

10. The time-series operators respect the time variable. L2.gnp refers to gnp_t-2 regardless of missing observations in the dataset. Notice in the following dataset that the observation for 1992 is missing:

+------------------------+ | year gnp L2.gnp | |------------------------| 1. | 1989 5452.8 . | 2. | 1990 5764.9 . | 3. | 1991 5932.4 5452.8 | 4. | 1993 6560.0 5932.4 | <- Note, filled in correctly 5. | 1994 6922.4 . | 6. | 1995 7237.5 6560.0 | +------------------------+

11. Time-series operators work with panel data as well as pure time-series data. The only difference is in how you tsset your data.

. webuse grunfeld . tsset company time

12. The time-series operators L. and F. may be combined with factor variables.

. sysuse census . generate t = _n . tsset t . regress death medage iL.region [aw=pop]

You could have specified Li.region instead of iL.region; the results would be the same.


. webuse gxmpl1 . list cpi L.cpi L2.cpi L3.cpi . list cpi L(1/3).cpi

. list cpi L2.cpi LL.cpi

. list cpi F.cpi F2.cpi F3.cpi . list cpi F2.cpi FF.cpi

. generate cpi_diff = cpi[_n] - cpi[_n-1] . list cpi cpi_diff D.cpi

. list cpi D.cpi S.cpi D2.cpi S2.cpi

Video example

Time series, part 3: Time-series operators

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index