Creating individual ID in a household survey dataset

[Zuhumnan Dapel]

Dear Users,
I've got household survey dataset with information on each household member including the household heads. The dataset has not individual ID but it has household. In the setting of the dataset, the first person in the household is the head. How do I create a variable of individual ID such that all the household heads with take the value of 1?
Thank you,
Dapel

[Svend Juul]

If the household ID is hhid and you want to generate pid as the individual ID within each household, first sort by hhid with the stable option to preserve the sequence within the household:

sort hhid , stable
by hhid: generate pid = _n

Now number 1 in each household is the head.

[Zuhumnan Dapel]

This is beautiful. It works perfectly. Thanks a million dear Svend. How do I create the variable for the household size in each household?

[Svend Juul]

sort hhid pid
by hhid: generate hhsize=_N

* About _n and _N, see help subscripting

[Zuhumnan Dapel]

This is really helpful. Thanks

[Stephen Jenkins]

Be very careful regarding precision issues when creating ID variables. (There are previous posts about this.) If there are many respondents -- hence large numbers for ID -- then the default float format of generate may not be sufficient. Consider generate long IDVAR = .... Or hold IDs using a string variable. (See the previous posts.)

[Zuhumnan Dapel]

Dear Prof Jenkins, I've got over six hundred thousand observations.

[Clyde Schechter]

Stephen Jenkins is right that, in general, individual ID variables should be stored as longs or strings. In my proposed code, however, I deliberately used int because what we are generating here is a variable that counts up the number of individuals in a single household. For this purpose even -byte- would probably have been sufficient. And I was trying to be mindful of not wasting storage since I know that household survey data tends to include hundreds of thousands or millions of observations.