Hello everyone,
I was hoping to get a piece of clarification on the answer that's expected
to question 3 of the homework. The question prompts us to, "comment on
what [we] see" in the histogram and boxplot of variables "AGE" and "WARD."
For those of us without much statistical training, what type of comments
are expected? I understand from some of my more statistically adept
colleagues that there are some interesting features with respect to
"skewed distributions" and the like, but if one is not comfortable with
(or familiar with) this terminology, what is expected?
Thanks,
Lanhee
Some thoughts:
1) Did I mention that this class is a lot of work? ;-)
2) Remember that Tao is your buddy. It seems that we are moving to a world in
which sections are really all about working on the problem sets. I think
that this a probably a good thing, but Tao and I are working on the
details. As always, your thoughts and opinions are appreciated.
3) Remember that your buddy is your buddy. I hope that your study groups are
working out. Study groups are mandatory. So, even if *you* don't need your
study group, your study group needs you. Any study group problems
(especially any study group that is lost at sea) should be brought to my
attention.
4) One of the many mottos for GOV 1000: From each according to his ability, to
each according to his need. I have asked Tao to provide special attention in
office hours and sections -- which may soon be more or less the same thing
-- to those students/study groups who are having the most trouble. I want
Tao to spend 90% of his time on this. Others should make better use of our
wonderful class mailing list.
5) Today looks to be a beautiful day for lunch. Aren't you sad that you didn't
sign up? I still have openings for next Wednesday and Friday. Please e-mail
me.
6) I would like to modify the reading assignment for Monday, assuming that not
everyone didn't finish the readings last night. ;-) Let us postpone Neter
Chapter 1 till the following week. Instead, please read the handout for QR
33 that is now on the web site. The basic purpose of this week's reading is
to begin thinking about statistics as an experiment. Once that is clear, we
can start doing statistics on data that come from something other than an
experiment.
7) Note that I will assume that you have done the readings. In fact, early in
the lecture I will say, "Does anyone have any questions on the QR 33
handouts?" If you haven't read them, you won't have many questions.
8) The QR 33 handout is essentially all the handouts for, you guessed it, QR 33
a Core Undergraduate course taught by Don Rubin. It is the most accessible
collection of readings that I can find that show the "right" way to think
about statistics. In the following weeks, we will read more formal
presentations of this approach, but I *think* -- comments always welcome --
that this is the best way to get started. Special thanks to Don Rubin and
Samantha Cook (TF for QR 33) for making this material available to us.
9) Class will start at 2:07 on Monday and finish at 3:45. There is a seminar in
that roon directly after our class and, as a courtesy to them (they need to
rearrange all the tables) we will end early. I know that you will all be
deeply saddened at losing those few moments of scintillating lecture, but
please be brave.
Dave
--
David Kane
Lecturer In Government
617-563-0122
dkane(a)latte.harvard.edu
A (different) student writes:
>
> hey dave,
>
> since you've made the grammatical mistake twice (once in a
> previous email), it should be "your headache" not "you're headache" below.
>
If these are the only grammatical mistakes that you can find, then you
need to get more sleep!
;-)
The only thing worse than my grammar is my spelling!
Dave
> On Mon, 30 Sep 2002, Dave Kane wrote:
>
> > A student in the class writes:
> > > Dave,
> > >
> > > I am not sending this to the list because I don't want to seem like an
> > > idiot.
> >
> > You don't seem like an idiot to me. In fact, I have some of these
> > questions myself! Hope you don't mind that I sent this anonymously to
> > the whole list, but I am sure that others feel the same. I am still
> > conflicted about this reading assignment. On the one hand, it is still
> > the best simple introduction to these ideas that I can find. On the
> > other, these are clearly class notes and a lot of what is said in
> > class is missing.
> >
> > For now, I would recommend bringing the handout to class. We will make
> > time for questions and answers on it.
> >
> > > The QR 33 reading is less than clear for someone who has not seen this kind
> > > of notation before. Here are a few of the terms that I couldn't understand
> > > (and I'm not even very far into the reading yet):
> > >
> > > Example I-5: "Stability assumption"? What's that?
> >
> > Means that what is going on with one unit does not affect the other
> > units. The only place that this is violated in the handout is in the
> > first two pages where you're headache can make mine worse.
> >
> > > Example I-6: How does the doctor know what's best? It seems strange to have
> > > the outcome, but not the conditions specifying the outcome. Is the doctor a
> > > black box?
> >
> > It's a thought experiment. Highlights that when people have intuition
> > about how different treatments will work, then their assignments make
> > it difficult to evaluate which treatment is best.
> >
> > > Section I-3.1: What is the point of "Lord's Paradox"?
> >
> > To highlight that questions of causal effect are nonsensical without
> > clear thinking about treatments and controls.
> >
> > > Section I-4.3: How do I read/interpret that summation?
> >
> > That the only possible assignments have exactly "n" units get the treatment.
> >
> > > Section I-5.1: How do I read/interpret the pi notation?
> >
> > Mutiply everything together (similar to summation meaning add
> > everything together).
> >
> > > That's as far as I've gotten. I don't expect you to answer these questions
> > > now,
> >
> > Why not? I won't be waking up for another 5 hours! ;-)
> >
> > > but I would really appreciate it if you could read some of the notation
> > > in English tomorrow and explain how the experiments were set up.
> >
> > Excellent idea! Please bring questions. In fact, I would recommend
> > that everyone bring their own copies so that we can look at it
> > together. Might also be handy to bring the Green and Gerber article,
> > since we will use that as an example on several points. There is no
> > more important reading for the entire semester than this one.
> >
> > > I'm just
> > > not seeing the connection between this stuff and the (basic) probability
> > > stuff I've seen before (and have probably mis-remembered).
> >
> > It is not easy.
> >
> > Dave
> > _______________________________________________
> > gov1760-l mailing list
> > gov1760-l(a)fas.harvard.edu
> > http://www.fas.harvard.edu/mailman/listinfo/gov1760-l
> >
>
--
David Kane
Lecturer in Government
617-563-0122
dkane(a)latte.harvard.edu
A student in the class writes:
> Dave,
>
> I am not sending this to the list because I don't want to seem like an
> idiot.
You don't seem like an idiot to me. In fact, I have some of these
questions myself! Hope you don't mind that I sent this anonymously to
the whole list, but I am sure that others feel the same. I am still
conflicted about this reading assignment. On the one hand, it is still
the best simple introduction to these ideas that I can find. On the
other, these are clearly class notes and a lot of what is said in
class is missing.
For now, I would recommend bringing the handout to class. We will make
time for questions and answers on it.
> The QR 33 reading is less than clear for someone who has not seen this kind
> of notation before. Here are a few of the terms that I couldn't understand
> (and I'm not even very far into the reading yet):
>
> Example I-5: "Stability assumption"? What's that?
Means that what is going on with one unit does not affect the other
units. The only place that this is violated in the handout is in the
first two pages where you're headache can make mine worse.
> Example I-6: How does the doctor know what's best? It seems strange to have
> the outcome, but not the conditions specifying the outcome. Is the doctor a
> black box?
It's a thought experiment. Highlights that when people have intuition
about how different treatments will work, then their assignments make
it difficult to evaluate which treatment is best.
> Section I-3.1: What is the point of "Lord's Paradox"?
To highlight that questions of causal effect are nonsensical without
clear thinking about treatments and controls.
> Section I-4.3: How do I read/interpret that summation?
That the only possible assignments have exactly "n" units get the treatment.
> Section I-5.1: How do I read/interpret the pi notation?
Mutiply everything together (similar to summation meaning add
everything together).
> That's as far as I've gotten. I don't expect you to answer these questions
> now,
Why not? I won't be waking up for another 5 hours! ;-)
> but I would really appreciate it if you could read some of the notation
> in English tomorrow and explain how the experiments were set up.
Excellent idea! Please bring questions. In fact, I would recommend
that everyone bring their own copies so that we can look at it
together. Might also be handy to bring the Green and Gerber article,
since we will use that as an example on several points. There is no
more important reading for the entire semester than this one.
> I'm just
> not seeing the connection between this stuff and the (basic) probability
> stuff I've seen before (and have probably mis-remembered).
It is not easy.
Dave
I hope that everyone had a fun week-end.
1) There is a tradition where I come from that boring classes should
start with a joke. So, I'll ask if anyone has a joke at the start of
class tomorrow. I hope that someone will have one. Ideally, it would
be something about Harvard or Government or both, but any attempt at
humor would be appreciated.
2) I will be asking for volunteers tomorrow. Don't worry, it will be
painless and, I hope, amusing. At the least I can promise that it
won't be like any other Ph.D. class.
3) We will be spending quite a bit of class time on the QR 33 related
readings. (I (like to) think that the experimental design readings are
fairly clear.) I hope that you will read it closely and come with
questions.
Dave
--
David Kane
Lecturer in Government
617-563-0122
dkane(a)latte.harvard.edu
Subject: [gov1000] mistake in readings (?)
In-Reply-To: <003701c2673c$1aeaf660$4169f78c@hewlett2ih5nie>
References: <003701c2673c$1aeaf660$4169f78c@hewlett2ih5nie>
X-Mailer: VM 7.00 under 21.4 (patch 6) "Common Lisp" XEmacs Lucid
FCC: /home/login/dkane/vm-folders/sent
--text follows this line--
I concur.
I have cc'ed Professor De Veaux. I know he is eager to receive any
similar corrections or suggestions.
Thus do we enter the community of scholars.
Dave
Stanislav Markus writes:
> just in case somebody else is wasting the gorgeous day on the readings:
>
> could there be a mistake in DeVeaux ch. 2, in the last sentence of page
> 13? i think it should be:
> "The test statistic tells us that our estimate, 23.00, is 3.7884 [not
> 2.504] standard errors to the left of the hypothesized mean of 20 [not
> 23]."
>
> 2.504 is the sample standard deviation, and 23 is the actual estimate...
>
> cheers,
> s t a n
>
> _______________________________________________
> gov1760-l mailing list
> gov1760-l(a)fas.harvard.edu
> http://www.fas.harvard.edu/mailman/listinfo/gov1760-l
--
David Kane
Lecturer in Government
617-563-0122
dkane(a)latte.harvard.edu
just in case somebody else is wasting the gorgeous day on the readings:
could there be a mistake in DeVeaux ch. 2, in the last sentence of page
13? i think it should be:
"The test statistic tells us that our estimate, 23.00, is 3.7884 [not
2.504] standard errors to the left of the hypothesized mean of 20 [not
23]."
2.504 is the sample standard deviation, and 23 is the actual estimate...
cheers,
s t a n