From Manual:
sample p ( loadstatement | selectstatement )
where:
p is an arbitrary expression which valuates to a number larger than 0 and lower or equal to 1. The number indicates the probability for a given record to be read.
Example:
SAMPLE (0.10) LOAD * FROM Table; (This statement would load about 10% of the records in "Table")
I'm attaching screenshots from a QV document that autogenerates 1,000,000 numbers in two tables. One loads a sample of 10% the other a sample of 90%. You can see from my screenshots that in QlikView10 the probability is calculated incorrectly (looks like 1-p). The value "p" should be the probability that a record is read, not the probability that the record is skipped.
Table1:
SAMPLE 0.10
LOAD Rand() AS Random1
AUTOGENERATE 1000000;
Table2:
SAMPLE 0.90
LOAD Rand() AS Random2
AUTOGENERATE 1000000;