TINV returns a bad value in Excel. This is used for small sample s

[BHanl]

Try this spreadsheet in Excel 2003. It uses the TINV function.
T Score from inverse confidence and degrees of freedom. Uses two tailed
probabilities.
95% 0.05
97% 0.03
99% 0.01
Sample size 10
degrees of freedom 9
tscore 95% 2.262
tscore 97% 2.574 <== This is an impossible value.
tscore 99% 3.250 It must be between 2.62 & 3.25

There appears to be a serious bug in the TINV function. The content of the
tscore column C cells is: =TINV(C5,$C$8)
Try it. This doesn't work correctly. MS needs to fix it pronto. Hard to
notify them.

[Dav]

The t values look correct to me looking at other statistical tables. They are consistent with Excel so there is not a bug! Are you sure you are not misreading the 2.262 as 2.62. Obviously the value in the middle has to lie between 95% and 99% but I suspect you have misread the lower value, it is a Monday morning!

Regards

Dav

[BHanl]

Yes, I misread it. Should have quit earlier doing my verifications for each step. Had to do this for project up against SPSS. Glad it's my mistake. Good example of negative productivity.

[Jerry W. Lewis]

Excel 2000 provides 6-figure accuracy for the first two and 5-figure accuracy
for the third.

Excel 2003 provides 8-figure accuracy for the first and 9-figure accuracy
for the last two.

Jerry

"BHanl" wrote:

> Try this spreadsheet in Excel 2003. It uses the TINV function.
> T Score from inverse confidence and degrees of freedom. Uses two tailed
> probabilities.
> 95% 0.05
> 97% 0.03
> 99% 0.01
> Sample size 10
> degrees of freedom 9
> tscore 95% 2.262
> tscore 97% 2.574 <== This is an impossible value.
> tscore 99% 3.250 It must be between 2.62 & 3.25
>
> There appears to be a serious bug in the TINV function. The content of the
> tscore column C cells is: =TINV(C5,$C$8)
> Try it. This doesn't work correctly. MS needs to fix it pronto. Hard to
> notify them.