Does Analyze estimate statistics work?

From: Randy Fought <FOUGHTRL_at_CTRVAX.VANDERBILT.EDU>
Date: Mon, 27 Mar 1995 21:50:07 GMT
Message-ID: <1995Mar27.215007.3308_at_news.vanderbilt.edu>

        When I analyze a table based on a percentage of the rows instead of computing on the entire table I get odd values for the number of distinct values from the USER_TAB_COLUMNS view.

        For instance, when estimating statistics based on 25% of the rows USER_TAB_COLUMNS says:

	 column 1 has 48 values (48,132 when computing complete statistics)
	 column 2 has 38 values (instead of 365)
         column 3 has 38 values (instead of 235)
         column 4 has  2 values (this is correct)
	 column 5 has 10 values (this is correct)
	 column 6 has 38 values (instead of 5,133)
	 column 7 has  1 value  (this is correct)

	The fields with a small number of values are correctly

identified when statistics are estimated. Other columns are not even close. I understand that when you only look at every fourth tuple you will miss some values. All of the columns have a number of distinct values that ends in 8?

        The same types of values (numbers way too low and ending in eight) occur when other table's statistics are estimated.         

        My database is on an Intel Pentium with the NT OS. Any help would be appreciated.

        Randy Fought Received on Mon Mar 27 1995 - 23:50:07 CEST