The CONFIDENCE() function is rarely appropriate, because you rarely know the
population standard deviation. For an estimated standard deviation, you
should use
CONFIDENCE(alpha,S,n)*(-TINV(alpha,n-1)/NORMSINV(alpha/2))
The multiplicative factor which will roughly equal 1.25 when n=7

Jerry

"cwbecker" wrote:

The individual CONFIDENCE() function [CONFIDENCE(alpha, STDEV, n)] returns a
different confidence interval than if you use Descriptive Statistics Analysis
[ToolsData AnalysisDescriptive Statistics]. In fact, I've been witnessing
a 25% difference between the two values.
Any ideas why these two methods of calculating the confidence interval yield
different results?