Economics 761
Note that this course will not be offered in Fall 2016.
This page is for the course entitled Time Series Analysis. In fact, we
will not spend all our time on time series analysis, since the course is
intended to be an advanced topics course in econometrics, and just happens to
bear the name that it bears. It is more bureaucratic hassle than it is worth to
change the name every year to reflect the actual content of the course.
A class outline without much more information than what there is here is
found as a PDF file here.

Although the time slot for the course is Mondays, 11.3014.30, in Leacock 517, I
plan to meet on Tuesday evenings, 18.0021.00, in the same room.
The first class will then take place on Tuesday September 8.

My office is Leacock 507.

Those who wish can join me for a drink after class in Thomson House on upper
McTavish  the graduate students' club. Their selection of beer is quite wide,
although they seem not to have any noalcohol beer.
Software for models with AR(p) disturbances
This link is to a description of the
implementation of the artificial regression discussed in class for
regression models with AR(p) disturbances. The Ects code is
embedded in the discussion.
For actual use, the Ects code file is arml.ect.
The illustrative example is found here.
The data file is ar3.dat.
Useful papers:

This is the Maximum Likelihood by Artificial
Regression paper.

A survey of bootstrap methods written by James MacKinnon and me for a volume on
econometrics is here. It should be useful as
an introduction to the topic.

A more recent paper by me has appeared in Russian in the journal Quantile.
Here is the English version that was translated into
Russian. It has most of my latest thoughts on general bootstrap matters.

This link is to a set of notes on the
bootstrap originally prepared for a set of courses I gave in March and April 2012 in
Australia. The notes here were revised this year for presentations at CREATES in
Denmark. They are in the form of slides.

The algorithm for weighted resampling is given, along with C++ code, in this extract of the documentation of Ects.

The paper by Mirza Trokic and me on bootstrap iteration is
here.

The first discussion of the fast double bootstrap (FDB) is found in
a paper by MacKinnon and me on reliable bootstrapping.

The FDB works astonishingly well with the nonnested J test, as seen in
this paper.

A short survey by MacKinnon discusses other successes
of the FDB.

The wild bootstrap is the topic of this paper by
Flachaire and me.

Bootstrapping of nonlinear models can take advantage of artificial regressions,
as shown in this paper by MacKinnon and me.

Another example of bootstrapping that works well, in a purely linear context,
is found in this paper, in which we look at the DM
nonnested hypothesis test.

Bootstrapping with endogenous explanatory variables poses various special problems, and is tied
up with the issue of weak instruments. In this paper MacKinnon and I
give a fairly complete analysis of a simple case.

This recent paper discusses the power of bootstrap
tests, and advances the theory of the bootstrap.

This link is to a set of notes that do not as
yet constitute a paper. The topic is simulating the bootstrap discrepancy. The
discussion shows explicitly how to use random numbers in all the simulation
procedures.

Another link to a set of notes. These are the
notes in French that deal with how Edgeworth expansions can be applied to the
study of the bootstrap discrepancy. There is also a treatment of the problem
we looked at briefly in class, dealing with a ttest for the
expectation of a distribution to be zero, complete with the stochastic
expansion.

Here, in French, is a Master's thesis by one of my students in
Marseille, in which methods of estimating the parameters and covariance matrix of
estimates of the GARCH(1,1) model are discussed. Different initialisations are shown to have
little effect if the sample size is large enough.

For more about the Kalman filter, here is an introductory paper by
Welch and Bishop. I believe that
these are notes for a course they gave somewhere.

Next, a more complete discussion
by the same authors.

This link is to an extract from notes by Steven Shreve,
on which his finance textbook is based. The extract explains the ItÃ´ integral.

Here are some notes on the construction of a
sequence of stochastic processes on [0,1] that converge almost surely to
a standard Wiener process, thereby giving an almost sure version of the
functional central limit theorem.

The material on threestage least squares when only one equation of the system is
overidentified can be found in these notes.

Here is an old paper on C(alpha) tests.

The whole business of estimating the bootstrap discrepancy is discussed in Sections 5 and 6 of
this paper, in the context of bootstrapping a unitroot test.

This reference is not to an article
or a working paper, but to Luc Devroye's site, from which you can obtain what I called the
"bible" of generating realisations from all sorts of distributions, continuous and
discrete, from U(0,1) random numbers. The whole book can be downloaded for free.
In order to encourage the use of the Linux
operating system, here is a link to an article
by James MacKinnon, in which he gives valuable information about what software
is appropriate for the various tasks econometricians wish to undertake.
To send me email, click here or write directly to
Russell.Davidson@mcgill.ca.
URL: http://russelldavidson.arts.mcgill.ca/e761