Hi,
The first assignment is due this Wednesday in section. In addition to
your answers to the questions, you should also turn in your computer
codes. Please also indicate the names of other students you worked with.
Again, you are welcome to work together, but you should write up your own
answers. Good luck!
Kosuke
P.S. You should also find your partner for the term paper project by the
end of this week. If you can't find anyone, just e-mail me.
Yes, but is the "generic" logit function the one where beta = 1? i.e., if we
look at p21 of the lecture notes, #2 is logit, right? If you vary beta, you
get plots that look quite different. i.e., beta = 900 basically gives you a
step function, while beta = 0.001 gives you something that is almost linear.
What is the intuition behind choosing beta = 1, or am I just being wretchedly
wrong-headed?
Thanks,
Phillip.
-------------------------------------------------
Phillip Y. Lipscy
Perkins Hall Room #129
35 Oxford Street
Cambridge, MA 02138
(617)493-4893 DORM
(617)851-8220 CELL
lipscy(a)fas.harvard.edu
http://www.people.fas.harvard.edu/~lipscy/
First Year Student, Ph.D. Program
Harvard University, FAS, Department of Government
-------------------------------------------------
Quoting Kosuke Imai <kimai(a)fas.harvard.edu>:
> Remember that we are not doing the regression. y=logit(x) maps x, which
> varies from -infty to infty, into y, which ranges from 0 to 1. You need to
> plot this function by letting x vary, say, from -4 to 4. Do the same thing
> and compare it with the probit function.
>
> Kosuke
>
> On Tue, 11 Feb 2003, Phillip Y. Lipscy wrote:
>
> > For the logit function, are we assuming that beta = 1 or is this supposed
> to
> > be intuitively obvious? Thanks,
> >
> > Phillip.
> >
> > -------------------------------------------------
> > Phillip Y. Lipscy
> > Perkins Hall Room #129
> > 35 Oxford Street
> > Cambridge, MA 02138
> > (617)493-4893 DORM
> > (617)851-8220 CELL
> > lipscy(a)fas.harvard.edu
> > http://www.people.fas.harvard.edu/~lipscy/
> >
> > First Year Student, Ph.D. Program
> > Harvard University, FAS, Department of Government
> > -------------------------------------------------
> >
> >
> >
> >
> > _______________________________________________
> > gov2001-l mailing list
> > gov2001-l(a)fas.harvard.edu
> > http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
> >
>
> _______________________________________________
> gov2001-l mailing list
> gov2001-l(a)fas.harvard.edu
> http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
>
For the logit function, are we assuming that beta = 1 or is this supposed to
be intuitively obvious? Thanks,
Phillip.
-------------------------------------------------
Phillip Y. Lipscy
Perkins Hall Room #129
35 Oxford Street
Cambridge, MA 02138
(617)493-4893 DORM
(617)851-8220 CELL
lipscy(a)fas.harvard.edu
http://www.people.fas.harvard.edu/~lipscy/
First Year Student, Ph.D. Program
Harvard University, FAS, Department of Government
-------------------------------------------------
I'm not quite sure how to generate this Monte Carlo data set. Are we
suppose to just plug randomly generated numbers into the x_1s?
Also 2 states that Beta=(2,1). Are these the entries in the beta matrix?
Thanks a lot.
Andrew
How do I plot two things (say a and b) over each other? I know Kosuke
mentioned this in class but I have forgotten.
Thanks,
Andrew
--
Andrew Reeves
reeves(a)fas.harvard.edu
617.493.3485 tel.
301.639.8369 cell.
http://people.fas.harvard.edu/~reeves
If you calculate 95% CIs 1000 times, 950 of them should cover the true
value. Does this make sense?
Kosuke
> Well, each time we calculate a beta, we get a confidence interval. So
> at the end, we have many many confidence intervals for same number of
> betas. Which CI do I use? Do I just pick one and do it or take the
> average of all boundaries, etc?
>
>
"appropriate covarage probability" means, for example, your 95% confidence
interval covers the true value exactly 95% of time. If your 95% CI covers
the true value only 80% of time, then it's too short. If it covers 100% of
time, it's too wide!
Kosuke
> Hi Kosuke,
>
> The homework question 2 part c asks us to calculate many confidence
> intervals for beta hats. How would we show that these confidence
> intervals have appropriate coverage probability? I don't think I'm
> understanding the question right. Thanks!
>
Hello,
My name is Chester Lee and I'm a junior majoring in statistics.
I'm looking for a PhD student(s) to coauthor a paper. I've taken a lot
of undergraduate and graduate statistics courses over the years and
have been using R all along. I'm very interested in applying statistics
in social science, and that's why I'm taking this class. I'm willing to
work in any field, and I guarantee that you'll be pleased with my
analysis.
Here are some of the statistics classes I've taken that might be
relevant to your project.
Stat 111 - Theoretical statistics (maximum likelihood inference, etc)
Stat 139/Stat 149 - Regression Analysis (linear regression, logistic
regression, etc)
Stat 160 - Survey methods (everything you need to know about surveys)
Stat 220 - Bayesian Data Analysis (intense R programming)
Stat 214 - Causal Inference in Social and Biomedical Sciences (studying
causal relationships - also intense R programming)
Stat 210 - Probability
If you are interested, please email me.
Thanks,
Chester Lee
Hi everyone,
Do any of you plan on meeting tomorrow to do the homework? If you do,
could I join your study group? Just name the time and place and I'll be
there...
Thanks!
Maria
PS. I took Gov 1000 several years ago, so I have no experience with R
