Local expression? Trying to understand user-defined abar-command

[Jesse Wursten]

I'm trying to understand the user-written abar-command which defines the panel test for serial correlation introduced by Arellano & Bond 1991 (going through Stata files is sometimes easier than trying to understand the initial paper). At multiple occasions, they use the following syntax

tempvar wtvar e0
gen double `wtvar'`=cond("`e(cmd)'"=="newey", "=","")'`e(wexp)' if `touse'
...
 `=cond("`id'" == "", "", "by `id' :")' egen double `ewi' = sum(`ewvar')

What has me puzzled is that the expression after gen double `wtvar' is a local (`=cond("`e(cmd)'"=="newey", "=","")'), but the local is not defined anywhere and it also seems a very strange name for a local?

I'm guessing it's some sort of ad-hoc conditional statement evaluation? But even then, if the condition it's false, the line reduces to gen double `wtvar' `e(wexp)' if `touse', which makes no sense to me? Is this perhaps some old Stata syntax? The program runs under version 7.0.

I've attached the complete ado file if that helps.

[Nick Cox]

The syntax

`= exp'

doesn't define a named local macro. It means

1. carry out the calculation defined by the expression exp

2. substitute the result here.

Here are two examples:

. di "`=2 + 2'"
4

. sysuse auto, clear
(1978 Automobile Data)

. su mpg

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
         mpg |         74     21.2973    5.785503         12         41

. di "coefficient of variation is `= r(sd)/r(mean)'" 
coefficient of variation is .2716543385281728

.

Documented at help macro.

The first example you cite could, I guess quickly, be rewritten this way:

if "`e(cmd)'" == "newey" gen double `wtvar' = `e(wexp)' if `touse'
else gen double `wtvar' `e(wexp)' if `touse'

and seems designed to work around some idiosyncratic feature of newey.

In this case I think I prefer the latter form.

[Jesse Wursten]

The syntax

[...]

Ah! Normal commands return e(wexp) as "= weightvar", perhaps the old version of newey returned e(wexp) as "weightvar" instead. Thanks!