Hello All,
Home Work #5 has been posted to the course website.
On Monday, I will begin discussing regression. It would be helpful if
people read ahead both in the lecture notes and the textbooks in
preparation for Monday's lecture.
Jacob will offer a review section on Wednesday. Unlike other
sections, this section will not be split and everyone should show up
around 6.
Cheers,
JS.
Hi~ all
I think it might be really helpful if Jacob will post answers to homework (2-4) on the
course website.
Although everybody knows the solution, the "standard" answer keys will be great for us
to prepare for the midterm!
It says they're independent?? How could I have missed that??!? Now I
feel dumb...
Thanks Ian!
Z
Ian Yohai wrote:
>Remember the question says that X1, X2, X3 are independent random variables,
>and hence the covariance terms drop out.
>
>Also I think you can always think of the sum of three variables as the sum
>of two variables...e.g., think of a variable Z = 2*X1 - X2, then you can
>take the variance of (Z + 4*X3), which is equal to the variance of Y.
>Essentially you just apply the formula for variance of the sum of two
>variables twice.
>
>Hope that helps,
>Ian
>
>-----Original Message-----
>From: gov1000-list-admin(a)fas.harvard.edu
>[mailto:gov1000-list-admin@fas.harvard.edu] On Behalf Of Zoë VanderWolk
>Sent: Tuesday, October 28, 2003 8:10 PM
>To: gov1000-list(a)fas.harvard.edu
>Subject: [gov1000-list] Those covariances...
>
>So who knows what to do with the variance of a sum of three random
>variables? Can covariance have three covariates within the brackets?
>ie Cov(X1, X2, X3)? And if so, what does the formula look like? Or do
>we have Cov(X1, X2) + Cov(X2, X3) + Cov(X1, X3)?
>
>Zoë
>
>
>
>_______________________________________________
>gov1000-list mailing list
>gov1000-list(a)fas.harvard.edu
>http://www.fas.harvard.edu/mailman/listinfo/gov1000-list
>
>
So who knows what to do with the variance of a sum of three random
variables? Can covariance have three covariates within the brackets?
ie Cov(X1, X2, X3)? And if so, what does the formula look like? Or do
we have Cov(X1, X2) + Cov(X2, X3) + Cov(X1, X3)?
Zoë
> What are the best place to learn more twin studies and the "DID"
> methodology ?
On twin studies, a good review is offered by the American Society of
Human Genetics statement entitled "Recent Developments in Human
Behavioral Genetics: Past Accomplishments and Future Directions". It
was published in _American Journal of Human Genetics_ 60:1265-1275,
1997. I've attached a pdf version of the article below.
The most relevant section of the article is entitled "twin and
adoption studies". In this section, identical and fraternal twin
studies are discussed---their assumptions, benefits and problems. The
article offers the example of schizophrenia for which identical
vs. fraternal twin studies nailed down that a big component of the
condition is genetic. The article outlines how the leverage comes
from the fact that the correlation between identical twins reflects
all of the genetic variance whereas that between fraternal twins
reflects only half of the (additive) genetic variance.
There is a large and growing literature in Difference in Difference
research designs. A relatively straightforward overview is offered
in:
@Article{
author = {Meyer, Bruce D.},
title = {Natural and Quasi-experiments in Economics},
journal = {Journal of Business and Economic Statistics},
year = 1995,
volume = 13,
number= 2,
pages = {151--161}
}
I don't have a PDF of this article.
The MIDTERM EXAM will cover material from day one of the class up to
and including sampling.
Cheers,
JS.
1) The class will start 10 to 15 minutes late as usual because the
previous class does not get out in time. Consequently, the class will
also end late.
2) I will cover sampling today (which starts on page 103 of my lecture
notes). This is covered efficiently in chapter 6 of Wonnacott and
Wonnacott. The subject is also covered in Freedman chapters 19-23.
The lecture notes contain more detail than the text books. Although
Freedman is rather verbose, it offers great intuition.
Cheers,
JS.
Alex,
> But in part a we didn't calculate intervals, we calculated values.
Those values define the intervals. In 3a you are asked to calculate
the 99, 95, and 80% intervals. That is, the values between which, for
example, 99 percent of the normal density with mean 0.05 and variance
26 should be. In part 3b you are asked to calculate what proportion
of changes in approval are actually within the bounds calculated in
3a.
See the bottom of http://jsekhon.fas.harvard.edu/gov1000/normal1.R for
an example.
JS.
Alexander Liebman writes:
> Hi,
>
> Quick question about 3b. It asks us to compare the percentage of actual
> changes in presidential approval with the "intervals we just calculated [in
> part a]." But in part a we didn't calculate intervals, we calculated values.
>
> I see how to do the problem by comparing it to what the percentages should
> be, so I just want to make sure that this is what we're supposed to do.
>
> Thoughts?
>
> Alex
>