Statistics regarding at least one attribute (or column) of a table in a database system are collected using an enhanced mechanism for faster collection. The statistics collection is based on scanning a sample of rows of a given table, with the sample being less than all the rows of the table. The percentage of rows to be read is specified in a statement submitted to the database system. The database system then uses the specified percentage to read the sample rows from the table for collecting statistics on the at least one attribute.