Explicit subscripting []

[Ariel Karlinsky]

The more I use stata, the more I understand that I don't understand explicit subscripting if it contains variables.
For example, the following code is very simple: it would generate the first difference (between adjacent, not necessarily consecutive times), for each person, in ln_wage:

webuse nlswork, clear
bysort idcode: gen d_lnwage = ln_wage[_n] - ln_wage[_n-1]

However, I sometime see code such as:

by serial: replace famunt2=famunit[HHhead]

And I don't understand what it's supposed to even do. The help file for subscripting doesn't contain any examples or explanations for something like this. Any guidance or explanation would be extremely helpful.

[Nick Cox]

There's no example here, but first Stata will evaluate the subscript expression. Just as

_n - 1

is evaluated as the current observation number MINUS 1, so also

HHhead

is evaluated as the value in the present observation. But then by: implies that (e.g.) 1 will mean the first observation in the by: group (here by context a set of observations for a given family. So, you can work it out that with data like

clear 
input serial HHhead 
1       1
1       1   
1       1  
2       1  
2       1   
2       1
2       1   
end

HHhead[1] will mean a cross-reference to observations 1 and 4 for serial 1 and 2 as they are the first observations within each panel.

A teacher's tip is that I only thought I understand by: fairly thoroughly when I'd written a paper explaining it!