In article <rcs8.121.009B09BF@psu.edu>, rcs8@psu.edu (Roger C. Shouse) writes:

>Subject: Re: How to delete a bunch of observations? >From: rcs8@psu.edu (Roger C. Shouse) >Date: Tue, 28 Jan 1997 13:48:27 UNDEFINED > >In article <5cldfr$2fg@charm.magnus.acs.ohio-state.edu> cai.23@osu.edu >(Liming Cai) writes: >>From: cai.23@osu.edu (Liming Cai) >>Subject: How to delete a bunch of observations? >>Date: 28 Jan 1997 17:36:59 GMT > >>Hi, > > >>I have a dataset of N countries, each over 30 years, from 1960 to 1989. One >of the interesting >>variables is GDP. However, for some countries, GDP was not available in >1960. What I would >>like to do is to delete those countries, their entire history of 30 years, >if they don't have a GDP data >>in 1960. Anybody can help me on that? Thank you very much. > >Depending on how you've specified your missing data, it may be as simple as: > >data <newdata>;set <olddata>; >if GDP = . then delete; > >assuming here that missing values are represented by '.' > >

That will delete observations where the GDP is missing. It will NOT delete all observations for a country where at least one observation for that country has a missing GDP.

There are a couple of ways to tackle the task. If you want to use procedural code, you could do something like:

PROC SORT DATA = COUNTRIES OUT = SORTED; BY COUNTRY GDP; DATA NOGDP (KEEP = COUNTRY); SET SORTED; BY COUNTRY; IF FIRST.COUNTRY AND GDP = .; DATA GDPONLY; MERGE COUNTRY (IN = A) NOGDP (IN = B); BY COUNTRY; IF A AND NOT B; RUN;

Alternatively, you could take an SQL approach:

PROC SQL; CREATE TABLE GDPONLY AS SELECT * FROM COUNTRIES WHERE COUNTRY NOT IN (SELECT DISTINCT COUNTRY FROM COUNTRIES WHERE GDP IS MISSING); QUIT;