Thanks, your advice worked. I don't have much experience with maths, and
therefore tried to stay away from dealing with optimization, but going
down to this level opens a lot of possibilities. For the record, the
code I used, as you suggested:
###
shape - mean(data)^2/var(data)
scale
I'm sorry, but I don't have time to read all your code. However,
I saw that you tested for x alpha in your Pareto distribution
example. Have you considered reparameterizing to estimate log.del =
log(alpha-min(x))? Pass log.del as part of the vector of parameters to
estimate, then
Thanks for the help, the wrapper function was very useful. I managed to
solve the problem using Spencer Graves' suggestion. I am analyzing the
interarrival times between HTTP packets on a campus network. The dataset
actually has more than 14 Million entries! It represents the traffic
generated by
Are you interested in turning that into a monitor, processing each
day's data sequentially or even each entry as it arrived? If yes, you
may wish to evaluate the Foundations of Monitoring documents
downloadable from www.prodsyse.com. If you have any questions about
that, I might be able
Dear R Users,
I am trying to obtain a best-fit analytic distribution for a dataset
with 11535459 entries. The data range in value from 1 to 3. I
use: fitdistr(data, gamma) to obtain mle's for the parameters.
I get the following error:
Error in optim(start, mylogfn, x = x, hessian =
In my experience, the most likely cause of this problem is that
optim may try to test nonpositive values for shape or scale. I avoid
this situation by programming the log(likelihood) in terms of log(shape)
and log(scale) as follows:
gammaLoglik -
+ function(x, logShape, logScale,
Spencer Graves's suggestion of using shape and scale parameters on a log
scale is a good one.
To do specifically what you want (check values for which the objective
function is called and see what happens) you can do the following
(untested!), which makes a local copy of dgamma that you
PS. 11 MILLION entries??
On Tue, 30 Sep 2003, Ben Bolker wrote:
Spencer Graves's suggestion of using shape and scale parameters on a log
scale is a good one.
To do specifically what you want (check values for which the objective
function is called and see what happens) you can