Gaurav,
The Anderson-Darling goodness of fit test is most commonly used to determine
if a sample of data fits a Normal distribution (or more accurately, it will
tell you if it is unlikely that it does not!). This array formula will
calculate the Anderson-Darling goodness of fit result (A^2) for a normal
distribution for a sample of data in named range "AD_Data":
=-ROWS(AD_Data)-SUM((ROW(OFFSET($A$1,0,0,ROWS(AD_Data),1))*2-1)*(LN(NORMDIST(SMALL(AD_Data,ROW(OFFSET($A$1,0,0,ROWS(AD_Data),1))),AVERAGE(AD_Data),STDEV(AD_Data),TRUE))+LN(1-NORMDIST(LARGE(AD_Data,ROW(OFFSET($A$1,0,0,ROWS(AD_Data),1))),AVERAGE(AD_Data),STDEV(AD_Data),TRUE))))/ROWS(AD_Data)
This is an array formula, so hold Ctrl+Shift as you enter it.
This formula gives the raw test result. You will most likely need to adjust
the result to compensate for sample size. You then need to compare it with
the critical test value for the alpha/power that you require. Different
sample size adjustments and critical values are published. More details here:
http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm
It is possible to obtain a p-value from the test result, but I have never
managed to find how this is done (not just a chisqr test etc.). Pehapse
someone else can assist here...
Cheers,
Dave