How to use proc sort nodupkey??

[OceanDream]

I have a dataset that has some duplicate observations as follows:

ID  VAR1 VAR2 VAR3

1      a        .          .

1      a        b        c

2      d        e         f

3      g        .         .

var1, var2 and var3 are numeric.

I want to delete the first observation. I use the following code:

Proc sort data=have out=want nodupkey;

by ID VAR1;

run;

However, this code delete the second observation, instead of the first one.

Any one knows what I should do to delete the first observations?? Thanks in advance!

[Ksharp · Accepted solution]

data want;
 set have;
 by ID VAR1;
 if last.VAR1;
run;

Assuming your data records are sorted in a sequence where you want to keep just the last record with the key value.

See the note from @LinusH:

@Ksharp's code works if your rule is "in case of duplicates remove the first observation".
But your sample data illustrates rather a rule that says remove the observation that has missing values on the non id variables. If this is the case, you need to presort your data to be sure to get the desired result. 

[OceanDream]

Thank you! Keshan!

[LinusH]

@Ksharp's code works if your rule is "in case of duplictes remove the first observation".
But your sample data illustrates rather a rule that says remove the observation that has missing values on the non id variables. If this is the case, you need to presort your data to be sure to get the desired result.