2 messages · Yvonnick Noel, Martin Maechler
I am trying to write a function that will solve a simple system of nonlinear equations for the parameters that describe the beta distribution (a,b) given the mean and variance. mean = a/(a+b) variance = (a*b)/(((a+b)2) * (a+b+1)) Any help as to where to start would be welcome.
The simplest way to go is probably to use the fitdistr function in MASS package. YNOEL
"YNoel" == NOEL Yvonnick <yvonnick.noel at uhb.fr>
on Thu, 01 Dec 2005 14:42:44 +0100 writes:
>>> I am trying to write a function that will solve a simple system of
>>> nonlinear equations for the parameters that describe the beta
>>> distribution (a,b) given the mean and variance.
>>>
>>>
>>> mean = a/(a+b)
>>> variance = (a*b)/(((a+b)2) * (a+b+1))
>>>
>>> Any help as to where to start would be welcome.
>>
>>
YNoel> The simplest way to go is probably to use the fitdistr function in MASS
YNoel> package.
yes, definitely for estimating the beta parameters - namely do
something smarter than the above moment estimates.
However, I (and probably most respondents to this thread)
believe that the question was more general (as the subject
indicates) about solving non-lin eq systems, rather than about
estimating the beta parameters (by a sub-optimal method).
Martin Maechler, ETH Zurich