Compatible SQL function for Excel FDist - excel

Question

Does anyone know that is there a compatible function in SQL for Excel FDIST and FINV? If there is no, anyone has any idea how to build that? May be in C#?
Thanks. Your help is greatly appreciated.

Answer 1

I have managed to resolve my problems by using a library from .Net Framework 4.0 and above (System.Windows.Forms.DataVisualization.Charting.StatisticFormula).
I am able to develop a function in C# using the above library for my calculation process. This is a powerful library where you can find mostly common use statistical formula in there (e.g. mean, median, t distribution, f distribution, and inverse of them.)
Below are the code snippet from me:
using System.Windows.Forms.DataVisualization.Charting;
private Chart ch = new Chart(); // You will need to declare an object of Chart type, as Statistic Formula class does not have a public constructor
double fDist = ch.DataManipulator.Statistics.FDistribution(fRatioVariance, degreeFreedom1, degreeFreedom2);
Hope this will help others. Thanks.

Answer 2

Though its too late but below are some statistical function implementation in SQL Server itself.
To get FDist function (equivalent to Excel - FDist), we will be needing Complete and Incomplete beta function as well as gamma function:
--GAMMA Function
CREATE FUNCTION [dbo].[udf_Gamma]
(
#x Float=NULL
)
RETURNS Float
AS
BEGIN
Declare #f Float = 10E99;
Declare #g Float = 1;
if ( #x > 0 )
Begin
while (#x < 3)
Begin
SET #g = #g * #x;
SET #x = #x + 1;
End
SET #f = (1 - (2/(7*power(#x,2))) * (1 - 2/(3*power(#x,2))))/(30*power(#x,2));
SET #f = (1-#f)/(12*#x) + #x*(log(#x)-1);
SET #f = (exp(#f)/#g)*power(2*PI()/#x,0.5);
End
else
Begin
SET #f = 10E99
End
return #f;
END
--BETA Complete Function
CREATE FUNCTION [dbo].[udf_BetaC]
(
#x Float=NULL
,#a Float=NULL
,#b Float=NULL
)
RETURNS Float
AS
BEGIN
--double betacf(double a,double b,double x){
Declare #maxIterations int = 50, #m int =1
Declare #eps Float = 3E-5
Declare #am Float = 1;
Declare #bm Float = 1;
Declare #az Float = 1;
Declare #qab Float = #a+#b;
Declare #qap Float = #a+1;
Declare #qam Float = #a-1;
Declare #bz Float = 1 - #qab*#x/#qap;
Declare #aold Float = 0;
Declare #em Float, #tem Float, #d Float, #ap Float, #bp Float, #app Float, #bpp Float;
while((#m<#maxIterations) AND (abs(#az-#aold)>=#eps*abs(#az)))
Begin
SET #em = #m;
SET #tem = #em+#em;
SET #d = #em*(#b-#m)*#x/((#qam + #tem)*(#a+#tem));
SET #ap = #az+#d*#am;
SET #bp = #bz+#d*#bm;
SET #d = -(#a+#em)*(#qab+#em)*#x/((#a+#tem)*(#qap+#tem));
SET #app = #ap+#d*#az;
SET #bpp = #bp+#d*#bz;
SET #aold = #az;
SET #am = #ap/#bpp;
SET #bm = #bp/#bpp;
SET #az = #app/#bpp;
SET #bz = 1;
SET #m = #m + 1;
End
return #az
END
--BETA INCOMPLETE Function
CREATE FUNCTION [dbo].[udf_BetaI]
(
#x Float=null
,#a Float=null
,#b Float=null
)
RETURNS Float
AS
BEGIN
Declare #bt Float=0.0
Declare #beta Float=0.0
if( #x=0 OR #x=1 )
Begin
SET #bt = 0
End
else if((#x>0) AND (#x<1))
Begin
SET #bt = (Select dbo.UDF_Gamma(#a+#b)* power(#x,#a)* power(1-#x,#b)/(dbo.UDF_Gamma(#a)*dbo.UDF_Gamma(#b)) )
End
if(#x<(#a+1)/(#a+#b+2))
Begin
SET #beta = (Select #bt*dbo.udf_betaC(#x,#a,#b)/#a)
End
else
Begin
SET #beta = (Select 1-#bt*dbo.udf_betaC(1-#x,#b,#a)/#b)
End
Return #beta
END
--FDist Function
CREATE FUNCTION [dbo].[udf_FDist]
(
#x Float=NULL
,#df1 Float=NULL
,#df2 Float=NULL
)
RETURNS Float
AS
BEGIN
Declare #x1 Float=(#x*#df1)/((#x*#df1)+#df2)
return (select 1 - dbo.udf_BetaI(#x1,(#df1/2),(#df2/2)))
END
Check in Excel =FDIST(0.5,1,1)=0.608173448
and in SQL editor = SELECT udf_FDIST(0.5,1,1)=0.608173457369209
Regards,
Avi

Answer 3

FDIST an FINV Don't exists IN Sql Server.
You can write a SQL Sever function to implement two excel feature.
Show here to create function

Answer 4

in MS SQL Server there is standard_deviation
and some more statistical functions including variance.
There is also a million workarounds, for example: http://oreilly.com/catalog/transqlcook/chapter/ch08.html
,you need to dig deep into the math for these, though.