Statistics 8913 (Geyer, Spring 2020)

Announcements

New! The right way to make absolute value and norm signs in LaTeX is \lvert x \rvert and \lVert x \rVert. The first does |x| and the second does ||x||, but LaTeX looks better than HTML.

If one wants to make commands for these do

\newcommand{\abs}[1]{\lvert #1 \rvert}
\newcommand{\norm}[1]{\lVert #1 \rVert}

in the preamble. Then one can say \abs{x} and \norm{x}.

New! And another thing about LaTeX. Operators are different. You should know to use \exp instead of exp, \log instead of log, \sin instead of sin, and so forth. The letters exp in math mode (exp) mean e times x times p, that is, e × x × p, not the exponential function; you need \exp for that.

Similarly for operators that are not built into LaTeX. You need to define the operators yourself and then use them. In the preamble of your document (between \documentclass{beamer} and \begin{document}, or whatever document class you are using) put something like

\usepackage{amsmath}
\DeclareMathOperator{\var}{var}
\DeclareMathOperator{\cov}{cov}
\DeclareMathOperator{\sgn}{sgn}

and use them, for example, \var(X) rather than var(X).

Sometimes you don't want a unary operator like \exp or \log. Sometimes you want a binary operator like + or × or a binary relation like ≈ or ≡ When LaTeX does not have them you can make them as follows

\newcommand{\reland}{\mathrel{\rm and}}
\newcommand{\relor}{\mathrel{\rm or}}
\newcommand{\opmod}{\mathop{\rm mod}}

Notice that we could not use the names \and, \or, and \mod for these commands because they already have definitions in LaTeX and we did not want to clobber those definitions by using \renewcommand. It is somewhat a matter of taste as to what is a binary operator and what is a binary relation. The ones that use \mathrel above could arguably use \mathop instead.

New! And another thing about LaTeX. There are only two places where you need to add extra space in math because LaTeX cannot figure it out for itself (LaTeX is generally smarter than you are about math layout, but not here). In integrals, thin space (\,) before differentials

  \int_0^\infty
  \frac{\lambda^\alpha}{\Gamma(\alpha)} x^{\alpha - 1} e^{- \lambda x} \, d x
  =
  \frac{\Gamma(\alpha)}{\lambda^\alpha}

In sets, thin space (\,) inside the curly brackets for "set builder" notation, but not for sets as lists

  \{\, f(x) : x > 0 \,\}
  \{\, \theta \in \Theta : l(\theta) \ge c_\alpha \,\}

but

  \{ \text{cabbage}, \text{sauerkraut}, \text{kimchi} \}
  \{ 0, 1, 2, \ldots \}

For the first of these, I define a new command in the preamble of all my LaTeX documents that have fancy math

  \newcommand{\set}[1]{\{\,#1\,\}}

after which I can write

  \set{f(x) : x > 0}
  \set{\theta \in \Theta : l(\theta) \ge c_\alpha}

which looks a little cleaner.

New! Another LaTeX complaint: the spacing between an equation and its "condition" is \qquad as in

  f_{\alpha, \lambda}(x) = \frac{\lambda^\alpha}{\Gamma(\alpha)} x^{\alpha - 1}
  e^{- \lambda x}, \qquad x > 0

New! Use big delimiters in LaTeX. Stuff like

\log\left( \frac{1}{n} \sum_{i = 1}^n e^{\langle y_i, \theta \rangle} \right)

The \left and /right make the following delimiters (in this case round brackets) big enough to enclose the enclosed formula. There is also /middle for middle delimiters, like

  E_\theta\left\{ g(X)^{2^\alpha} \middle| X = x \right\}

Also there are delimiter sizes where you choose the size rather than let LaTeX figure it out for itself. These are especially useful when you don't want to change from round to square to curly brackets, like

  E_\theta\bigl( g(X) \bigm| X = x \bigr)

It allows you to tell the brackets apart.

New! The Canvas discussion group for the poster session has been set up. It opens Monday, April 20, just after midnight. It closes Wednesday, May 6, just after noon. All posters should be posted by Thursday, April 24, just after noon. There seems to be no way to get Canvas to enforce the latter requirement, but I will check. I hope no one objects to answering questions until May 6. If there is objection, we can discuss. If every student has asked at least one question of every other student, then the instructor will ask some questions, and it will all be over. So try to get your questions in as early as possible.

New! The quote less is more is from 20th century architect Ludwig Mies van der Rohe one of the leaders in modernist architecture.

We don't know if there will be the departmental awards ceremony on April 23. If not I guess the posters will be done via chat on April 24.

Info about Zoom https://it.umn.edu/canvas-host-zoom-meeting-within-canvas.

Link to canvas web site added to navigation at left.

Starting March 20, 2020 class goes on-line. No in-person meetings until further notice.

Every student will give one talk and make one poster (possibly on the same subject but not necessarily). Students who have already had 2 semesters of this will do 40-45 minute talks. Others will do 20 minute talks (2 talks per class period).

We will have a poster session. There is also the departmental poster session for the Advisory Board. There is also a poster session about reproducibility (so says Galin but I haven't found that) if anyone wants to do a poster on that subject.

If you have your own research to talk about, you may talk about that. Otherwise, you may talk about anything involving statistics. Talk about any paper in a statistics journal, or any paper in a science journal that involves statistics in an interesting way.

Instructor's talk for the first day of class: PDF and LaTeX source.

Schedule for talks and posters.

Lines of the above table about poster session and IRSA conference are now wrong and have been removed.

Navigation

• University of Minnesota
• School of Statistics
• Stat 8913
  • Canvas

Contents

• Announcements