Assuming your data is in files on HDFS, my understanding is that you cannot really do arbitrary in-place updates like you would with SQL in a relational database. You would probably need to read the data and modify the relevant records before writing it all back to HDFS. If you're using Hive or HBase to store your data, then maybe there are other options available, but in-place updates are not really what Hadoop is intended for.
Pig is not a database in which you can update/delete data. It basically reads data from HDFS/local file system and does operations on it. So you should not try updating/deleting that data.