distribution curve

  • Thread starter Thread starter mark scholes
  • Start date Start date
M

mark scholes

Hi,
I have the strandard deviation, mean and skew of a curve
and I wish to draw the curve. Is there a function to do
this or do I have to write one. Where can I find a
function if someone else has already done so.

Thanks
MarkS
 
mark scholes said:
What more information do I need
Click to expand...

You either need to assume a functional form or you need a bunch of data
points to fit to likely functional forms, then select the 'best' one ('best'
is a loaded term in this instance).

The problem is that having mean, standard deviation and skewness is
sufficient to indentify a particular member of a 3-parameter family of
distributions, e.g., the Burr distribution, or an arbitrarily shifted member
of a 2-parameter family of distributions, but it's not sufficient to select
which distribution family to use.
 
If you are willing to assume that higher order cumulants are zero (or
nearly so), then you could use the Edgeworth expansion (equations
26.2.47 and 26.2.48 from Abramowitz and Stegun's Handbook of
Mathematical Functions, National Bureau of Standards and later Dover
paperbacks, with various printing years from 1964 into the 1970's)

Jerry
 
Do you have access to the original data? If so, you can toss the
summary stats, using the original data's empirical cumulative as the
best (MLE) proximal to the "true" distribution, sans additional
assumptions.

Good luck. Perhaps the person who posed this to the OP has a string
prior as to which family this belongs.

HTH
Dave Braden
 
Back
Top