Tcl Library Source Code

The documentation for math::statistics::test-normal says it returns 1 if the data is distributed normally within the given confidence.
I am not a statistician, but either I am severely confused or this is completely wrong.  It seems that this procedure returns 1 when the null hypothesis of a normal distribution should be rejected, which is the opposite of what I read.

Also, the confidence specification seems backwards from the things I read---which admittedly are just Wikipedia and MATLAB help---which use significance (alpha), a small number, rather than confidence.  In fact, I don't even think confidence is the right word here, or at least I don't think you're talking about confidence intervals.

package require math::statstics
0.6

I have adjusted the documentation and the implementation to use significance (as the complement to the original parameter). The new implementation is actually compatible with the old one, but emphasis is put on the new one.

"Confidence" is not the correct word indeed, it is not about confidence intervals. I have tried to get a clearer idea about how to perform the test, but it is confusing :(.

To be continued.

Hm, will have to study your remarks more closely. Thanks for at least noticing this.