Andrew,
We should have included a reference (which I'll make below), we went
over reparameterization in secton on thursday. The basic idea is
that optim looks over the whole real number line to find a value
that maximizes a function, but
there are often instances where we need to constrain the search. For
example, when we want to find the maximum likelihood estimate of a
probability, we only want optim to search between 0 and 1.
Reparameterization (using some function) allows us to
constrain the search without resorting to constrained optimization
routines, which tend to perform much more slowly than optim.
The video from section has been posted on the website, and we've included
an example of reperameterization in the section notes on the class
website. Note, that this example is for constrained search over the
positive part of the real line.
Cheers,
Justin
On Sat, 3 Mar 2007, Andrew Coe wrote:
Hey everyone,
Can someone help me to understand what the second part of problem 1.4 is
asking? I do not understand what is meant by: "reparameterize pi ... Using
any cumulative probability distribution."
Thanks,
Andrew
On 3/2/07 2:55 PM, "Justin Ryan Grimmer" <jgrimmer at fas.harvard.edu>
wrote:
Hey Everyone,
We need to make a brief addition to problem 1.5. The number of trials for
the binomial distribution is 10 (T=10). I'll post appropriate changes to
the problem set momentarily--
Cheers,
Justin
_______________________________________________
gov2001-l mailing list
gov2001-l at
lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l at
lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l