There's an article by GK (just Gary, not Gelman too) in the American
Journal of Political Science (30:3), entitled "How Not to Lie with
Statistics" that's pretty useful...it's on JSTOR.
Olivia.
Phillip Y. Lipscy writes:
> Quick question about Appendix I:
>
> Which of the following are you looking for?
> 1. Just the programs we used in R (this is my understanding from class)
> 2. "All the R code that you used to calculate the results you present"
>
> #2 is not identical to #1. Per our discussion in class, my understanding is
> that you don't want R code beyond the programs. If we gave you all the code we
> used to get "any number you present," this is not much different from giving you
> a lot of raw output.
It should be. In an exercise like this, I end up trying dozens of
models and cleaning rules and exploratory data analysis. We do *not*
want to see all the stuff that your tried. We do want to see the code
that would allow us, starting with the unzipped ASCII data, to
"replicate" your results. Some of this code will do nothing but clean
the data. Even though it produces no output, we want to see it since
we need it to produce your output. Other code will actually calculate
the numbers that you present (and or be helper functions for doing
so). This sort code code (matching code, lm() code and so on) should
also be presented. We should be able to generate every number (mean
result and confidence interval from matching, regression coefficient,
and so on) that you present for ourselves. This is what replication
means.
Of course, this means that your code will need to be clear, clean and
(most important) well documented. A typical comment might be:
## Regression code used to produce the coefficients and standard
## errors in table 2. Note that the input dataframe is generated by
## running clean.data().
> Also, I noticed from the midterms that several papers contain additional
> appendices for things that don't fit into the main article. Is this a good
> place to dump a lot of extraneous stuff that doesn't fit into the
> main body
No, it is not.
> or
> do you want us to stick to the 2 appendix limit?
We do. If something is not important enough to tell us about in the
body of the paper, then it isn't important enough to tell us about.
Good luck!
Dave
> Thanks,
> Phillip.
>
>
> -------------------------------------------------
> Phillip Y. Lipscy
> Perkins Hall Room #129
> 35 Oxford Street
> Cambridge, MA 02138
> (617)493-4893
> lipscy(a)fas.harvard.edu
>
> First Year Student, Ph.D. Program
> Harvard University, FAS, Department of Government
> -------------------------------------------------
>
>
>
>
>
--
David Kane
Lecturer in Government
617-563-0122
dkane(a)latte.harvard.edu
If we are taking Gov 2001, can we assume our course1 accounts will stay
open for spring semester? (I have some files that I wouldn't want to
lose.) Just checking.
Thanks,
Anna
Hi, guys.
If you have a lattice graphic, you can change the size of the labels
using:
par.strip.text = list(cex = 1.5)
in your xyplot().
Still working on the axes...
Olivia.
Nirmala sent a nice message to R-help yesterday. Within 12 hours, she had two
answers, the second from one of the core developers of R, Brian Ripley.
One of the many reasons that we choose R for GOV 1000 (and all future GOV
empirical classes) is that this level of support is simply unavaible for any
other statistical package.
And it is free.
In terms of constructive criticism, the main thing that I would change about
Nirmala's post is that it should have provided code that creates the sample
dataframe y, as Ripley did in his reply. It is also customary to have a
somewhat more informative subject.
There is much to be learned from R help. Indeed, in recent days, as you can see
by browsing the archives, there have been a flurry of questions and answers
relating to lattice. Anyone doing serious methodology work should consider
subscribing to the list (I recommend the digest form) and reading the questions
and answers that seem interesting.
Anytime you receive answers to a question on R-help (or, indeed, anywhere), you
should send a nice thankyou e-mail to whoever took the time to help you out. If
you get a bunch of answers (especially if some of not cc'ed to the list), it is
customary to provide a summary post to the list of the original problem and the
various solutions, nicely organized, that were suggested.
Dave
Message: 11
Date: Mon, 6 Jan 2003 20:19:58 -0800 (PST)
From: Nirmala Ravishankar <n_ravishankar(a)yahoo.com>
To: r-help(a)stat.math.ethz.ch
Subject: [R] plot()
I am an R novice trying to figure out plot().
Specifically, I am trying to plot the values of a
numeric variable V for a set of years (1970, 1974,
1976, 1978, 1980). How do I get R to label the years
I am plotting on the x-axis rather then some general
levels (1970, 1975, 1980.) Using as.character(year)
doesn't seem to help, and using as.factor(year)
generates steps insteads of dots.
Help will be most appreciated. I have listed the code
I have been using below:
> plot(y$year, y$V, type = "b")
> plot(as.character(y$year), y$V, type = "b")
> plot(as.character(y$year), y$V, type = "b")
Thanks,
NR
--__--__--
Message: 12
Date: Mon, 6 Jan 2003 23:36:44 -0500 (EST)
From: Yuelin Li <yuelin(a)mail.med.upenn.edu>
Reply-To: Yuelin Li <yuelin(a)mail.med.upenn.edu>
Subject: Re: [R] plot()
To: r-help(a)stat.math.ethz.ch, n_ravishankar(a)yahoo.com
try text(c("1970", "1978", "1990"), x=1:3, y=1:3) after you first
call plot(c(1, 3), c(1, 3), axes=F, type="n", xlab="", ylab="")
to set the plotting area.
Yuelin.
-- From: Nirmala Ravishankar <n_ravishankar(a)yahoo.com>
To: r-help(a)stat.math.ethz.ch
Subject: [R] plot()
Date: Mon, 6 Jan 2003 20:19:58 -0800 (PST)
I am an R novice trying to figure out plot().
Specifically, I am trying to plot the values of a
numeric variable V for a set of years (1970, 1974,
1976, 1978, 1980). How do I get R to label the years
I am plotting on the x-axis rather then some general
levels (1970, 1975, 1980.) Using as.character(year)
doesn't seem to help, and using as.factor(year)
generates steps insteads of dots.
Help will be most appreciated. I have listed the code
I have been using below:
> plot(y$year, y$V, type = "b")
> plot(as.character(y$year), y$V, type = "b")
> plot(as.character(y$year), y$V, type = "b")
Thanks,
NR
______________________________________________
R-help(a)stat.math.ethz.ch mailing list
http://www.stat.math.ethz.ch/mailman/listinfo/r-help
--__--__--
Message: 13
Date: Tue, 7 Jan 2003 07:53:36 +0000 (GMT)
From: <ripley(a)stats.ox.ac.uk>
To: Nirmala Ravishankar <n_ravishankar(a)yahoo.com>
cc: <r-help(a)stat.math.ethz.ch>
Subject: Re: [R] plot()
First, it is usual to make plots with generic scales rather than label
each of the x values used, especially when they are irregularly spaced (as
here). But if you want to do that, use axis() to create your own axis.
Here's a test
y <- data.frame(year=c(1970, 1974, 1976, 1978, 1980), V=rnorm(5))
attach(y)
plot(year, V, type = "b", xaxt="n")
axis(1, year, year)
detach()
--
David Kane
Lecturer In Government
617-563-0122
dkane(a)latte.harvard.edu
Please avoid sending me Word or PowerPoint attachments.
See http://www.fsf.org/philosophy/no-word-attachments.html
Phillip Y. Lipscy writes:
> I was just reading the instructions for the exam, and under 17., it states that
> "there are many elections in which neither the Democrat nor the Republican won."
> Unless I am missing something, the data doesn't seem to tell us much about this
> because the dpercs only give us % of the two party vote. One can take out the
> data preceding a 3rd party incumbency, but that's only 28 data points, or 0.2%
> of the data. Is that "many?" Is there another dataset that tells us something
> about this?
>
> Thanks,
> Phillip.
>
You are not missing anything. Perhaps "many" should have been "several" . . .
Dave
--
David Kane
Lecturer In Government
617-563-0122
dkane(a)latte.harvard.edu
Please avoid sending me Word or PowerPoint attachments.
See http://www.fsf.org/philosophy/no-word-attachments.html