#value! error on worksheetfunction.gammadist in excel VBA

Question

I'm writing a code to find a fixed point to the equation:

1/N = xf(x)/(1-F(x)),

where x=p(1+phi) and f and F are the pdf and cdf of the gamma distribution and N is a positive integer.

The right hand side is known to be increasing in x for the gamma distribution, so there must be a unique solution.

I wrote the function:

Function find_pdstar_gamma(alpha As Double, beta As Double, phi As Double, N As Integer) As Double

Dim plow As Double
Dim pmid As Double
Dim ptop As Double
Dim parg As Double
Dim f As Double
Dim Fbar As Double
Dim z As Double

plow = 0
ptop = 1000
Do
    pmid = (ptop + plow) / 2
    parg = (1 + phi) * pmid
    f = WorksheetFunction.GammaDist(parg, alpha, beta, False)
    Fbar = (1 - WorksheetFunction.GammaDist(parg, alpha, beta, True)) 
    z = parg * f / Fbar - 1 / N
    If z > 0 Then
        ptop = pmid
    Else
        plow = pmid
    End If
Loop Until ((ptop - plow) / ptop) < 1e-08
find_pdstar_gamma = (ptop + plow) / 2

End Function

I get a #value! error when I run the function. Any idea where my error is?

Answer 1

I'm guessing that you're getting a run-time error 6, (overflow) or run-time error 11 (division by 0). This can actually be misleading, because it is actually a division by zero involving a Double can be reported as an overflow.

The first thing to do is check to make sure that your operator precedence is correct for this line of code:

z = parg * f / Fbar - 1 / N

This is evaluated as

z = ((parg * f) / Fbar) - (1 / N)

This means that if Fbar is ever zero (or a double suitably small enough), you'll get an error here.

The second thing to do is validate your inputs or add an error handler. That way you'll be able to get a more meaningful indication as to what is wrong.

Answer 2

I am going to answer this question: "Any idea where my error is?" I really have no earthly idea, but I know how to narrow it down. In referring to getting a #value error -- it seems that you are using it as a worksheet function. That may be the intended application, but it makes it difficult to debug. Instead - run it in the immediate window. In the VBE type something like

?find_pdstar_gamma(1, 3, 0.5, 100)

(but with the values of alpha, beta, etc. which are giving you grief). When you do this -- computation will stop with the offending line highlighted (if you hit the debug button in the error msgbox). This tells you where the error is.