Are you certain that the values are as you reported, rather than formatted
values with hidden non-zero decimal values? Alternately, you could try the
array formula
=PERCENTILE(ROUND(data,0),0.25)
entered with Ctrl-Shift-Enter to ensure that you are actually using the
posted data. The discrepancy between your reported value of 216.645 and
Doug's reported value of 216.75 for the reported data in Excel 2007 is easily
within the uncertainty that could have been introduced by formatting. The
basis is not clear for your expectation in the follow-up post that the 25th
percentile should be zero.
AFAIK all versions of Excel (at least since v4) have used the same algorithm
for Percentile, which should give 216.75 for the reported data. Note however
that there is no universally accepted definition of a percentile from sample
data. Hyndman and Fan, 1996, "Sample Quantiles in Statistical Packages", The
American Statistician 50(4):361-365 discuss 9 different definitions and
reference some others. Excel uses Hyndman and Fan's 7th definition, which
considers the min and max to be the 0th and 100th percentiles and equally
spaces the quantiles corresponding to intervening observations. This is a
reasonable description of the sample, but is almost certainly biased as an
estimate of the underlying population quantiles.
Jerry
