Bioequivalence and Bioavailability Forum

Dear HS,

Thank you very much for your comparison of Dr. Russel Lenth's NctInv() in Excel with OO Calc and R.

Since Delta is a optional var in Dr. Russel Lenth's NctInv(), in case of data without NC variable, you can recode
NCtInv = Solve("tDummy", p, x, y)
as
'NCtInv = Solve("tDummy", p, x, y)
to omit the execution of this code, then NctInv() will work well in your spreadsheet.

The difference between Dr. Russel Lenth's NctInv() in Excel and R may be caused by the different results of BetaInv() function generated with Excel and R.
Since I didn't install R in my computer, here I can only compare BetaInv() function in Excel with that in SAS.

Excel 2003:
=BETAINV(0.001,2,4)=0.010101318359375

SAS 9.1:
=BETAINV(0.001,2,4)=0.0101017879

OO Calc 3.3
=BETAINV(0.001,2,4)=0.0101017878837378

Nevertheless, Dr. Russel Lenth's NctInv() is a highly useful UDF for our calculation, although there may be a very very very small bias. It should be noted this bias is not cause by the algorithms itself but by the Excel's BETAINV(). But the lucky thing is MS has improved the precision of BETAINV() from 0.0000003 in Excel 2003 to about 10^(-15) in Excel 2007. Excel statistical functions: BETAINV


Based on the result of R, It seems that BETAINV() in OO Calc has higher precision than that in Excel.

Dear HS, do you agree with me?

BTW: Today is our National Day (since 1949-Oct-1), the whole country is full of happiness. I can't help myself to give my best wishes to all of you and your famaily and your country.