confusion with lead and lag in time series operators

[Lukas Rudolph]

Dear Statalisters,

I don't quite understand where my logic is wrong in the following example, but I would be happy if you could point that out:

I want to create a lagged variable and regress on that.

In the helpfile of the "time series operators" it says "For instance, L.gnp refers to the lagged value of variable gnp."
So what I take from this is that if I want to run a model with a lagged dependent variable I go with "regress variable L.variable"

However, creating a variable with "L.variable" this actually contains the lead of this variable?

*minimal working example
***************************************
sysuse uslifeexp, clear

keep if year < 1903
keep year le
tsset year

gen lead_le = F.le // operator meaning: F. lead (x_t+1)
gen lag_le = L.le // operator meaning: L. lag (x_t-1)

list
**************
=> output

+-----------------------------------------+
| year le lead_le lag_le |
|------------------------------------------|
1. | 1900 47.3 49.1 . |
2. | 1901 49.1 51.5 47.3 |
3. | 1902 51.5 . 49.1 |
***************************************

In the output above, the variable created with L.le contains the values of le in t+1 [and not t-1 as I expected]. Vice versa the variable F.le contains values of le in t-1 [which is the lag I want to use].
So what I take from this is that if I want to regress a variable on its value in t-1 I would have to use "regress variable F.variable"?

Am I going wrong here?

Thanks and best

Lukas

[Clyde Schechter]

I think you're just misreading your output. First, let's put it all in a code block so that it's readable and properly aligned (you should do this whenever you post code or results):

. sysuse uslifeexp, clear
(U.S. life expectancy, 1900-1999)

. 
. keep if year < 1903
(97 observations deleted)

. keep year le

. tsset year
        time variable:  year, 1900 to 1902
                delta:  1 unit

. 
. gen lead_le = F.le // operator meaning: F. lead (x_t+1) 
(1 missing value generated)

. gen lag_le = L.le // operator meaning: L. lag (x_t-1)
(1 missing value generated)

. 
. list, noobs clean

    year     le   lead_le   lag_le  
    1900   47.3      49.1        .  
    1901   49.1      51.5     47.3  
    1902   51.5         .     49.1

Everything is fine. In year 1900, the value for lead_le is 49.1, which is indeed the 1901 value for le, and the lag_le value is missing, as it should be since 1900 is the first year. Similarly in 1901, the value of lead_le is 51.5, which is indeed the 1902 value for le, and lag_le's value is 47.3, which was the 1900 le value. And so on. It's all correct.