On Sat, 15 Jul 2006 03:31:32 -0400, Heman Hunters <hemanhunters@GMAIL.COM>
>Today I have two questions:
Then you really should have posted twice, so that there would be sparate
>1) I have a large dataset, containing people's social security numbers and
>their income. No I want to select e.g. two percent of the people in the
>table AT RANDOM, and change their income by 10 percent. How do I
PROC SURVEYSELECT was recommended (and by someone other than David! :-) ).
That's probably the best choice, but you may not have it available.
Somebody suggested the venerable K/N algorithm. That's good if you must
control precisely the size of the sample so that you get not one more nor
one less observation than you specify.
If approximately two percent is good enough, it can be as simple as:
>2) Depending on what income, and the social status (married, divorced,
>etc.) the people above have, I want to give them a certain score. I have
>(on paper) a list of which income interval and social status goes with
>which score. Now I want to take the income variable in the table in 1),
>compare with my "lookup table", and assign the apropriate score to evey
>person. How do I do this the best way? Do I use a lot of IF:s? Or can I
>create a table containing income intervals, social status and the
>corresponding score, and look up values in some clever way?