Greetings,
The latex process is changing the order of my equations and my
charts such that the equations come between charts (not the order that
I intended.) Does anyone have any suggestions as to how to retain the
order of the objects in the latex document?
Sheldon
So i'm slightly confused by what exactly you want us to do here. It seems
that you want us to draw from "the" distribution, but the location parameter
translated into this case is mu_i, which is distinct for each observation
even if we set the other parameters equal to the MLE estimates. Do you want
us to take use mu' = E(X)%*%Beta? Or is it something more complicated that
i'm missing?
btw, does anyone else think it is rather cool that gamma(0.5) = sqrt(pi)!!
Hi everyone,
Materials from section tonight have now been posted, including the section
6 handout, associated R code, problem set 4 solutions and the new problem
set 5, due next Thurs.
In addition, details about the replication assignment that is due on April
3, 2006, and the memo due on April 10, 2006, are available below.
**Please be sure to read this handout and follow the instructions to ensure that
the process goes smoothly**:
http://www.courses.fas.harvard.edu/~gov2001/Assignments/repassign.pdf
Have a good weekend...
Best,
Ian
A good question--and yes, for this problem set, you should use listwise
deletion--that is, simply deleting the small number of rows that do not
have complete information for the covariates of interest.
Best,
Dan
----
Ph.D. Student
Department of Government
Harvard University
Tutor, Currier House
dhopkins(a)fas.harvard.edu
http://www.danhopkins.org
On Mon, 20 Mar 2006, Michael Law wrote:
> Hi Dan and Ian,
>
> Wasn't sure if this was discussed before, but should we be performing a
> complete-case analysis given the missing data in two of the variables?
>
> Thanks,
>
> Mike
>
Hi:
I am a bit unsure about what the INC (incumbent presidential party)
variable in the dataset means. it is coded -1,1 what do these values
stand for?
Thanks
Thomas
I am a bit puzzled by the choice of reparameterization in the examples where
sigma-squared is replaced by gamma. If the goal is to make sure that
sigma-squared is not negative, couldn't that be achieved by just making the
parameter be sigma? I have the feeling I'm missing something.
-Michael
===============
Michael J. A. Berry
Data Miners, Inc.
+1 617 742 4252
Hi All,
With the replication deadline coming up, I am going to hold an extra set
of office hours tomorrow, Monday, March 20th from after class until 5 PM
in the main computer lab on the ground floor of CGIS North.
Best,
Dan
----
Ph.D. Student
Department of Government
Harvard University
Tutor, Currier House
dhopkins(a)fas.harvard.edu
http://www.danhopkins.org
Hi everyone,
A second correction to PS 5 has now been posted. In part 3(b) fix \nu to
be 5. This also affects part 3 (d) as well.
Thanks to the group who pointed this out.
Many apologies for the confusion...
Best,
Ian
Thanks for the group below for asking these questions.
I've posted a slightly revised problem set, clarifying which model you
should run for Problem 3. You should run the model from Problem 2(b),
except omit the interaction term (that is, include an intercept,
JULYECQ2, ADAACA, INC, and DEV3). Y remains DVOTE.
\nu is just like any other parameter and hence can be optimized. So in
part 3(b) you should optimize \nu as well.
Best,
Ian
---------- Forwarded message ----------
Date: Sat, 18 Mar 2006 14:37:31 -0500 (EST)
From: Vipin Narang <vnarang(a)fas.harvard.edu>
To: Ian Brett Yohai <yohai(a)fas.harvard.edu>
Subject: problem 3
Ian, a couple of us are wondering what you mean by 'reparametrize' such
that nu is >2. nu is just degrees of freedom right, so it's not a
parameter to be optimized but is just user-specified if i understand it
correctly? so should we just specify any nu>2 for 3b? also, which model do
you want us to run for problem 3--the full one (with the interaction term
etc) or the restricted one?
thanks,
vip (and becky and justin and clayton)