% \documentclass[serif]{beamer} % Serif for Computer Modern math font.
\documentclass[serif, handout]{beamer} % Handout mode to ignore pause statements
\hypersetup{colorlinks,linkcolor=,urlcolor=red}

\usefonttheme{serif} % Looks like Computer Modern for non-math text -- nice!
\setbeamertemplate{navigation symbols}{} % Supress navigation symbols at bottom
% \usetheme{Berlin} % Displays sections on top
% \usetheme{Warsaw} % Displays sections on top
\usetheme{Frankfurt}  % Displays sections on top: Fairly thin but swallows some material at bottom of crowded slides
\usepackage[english]{babel}
\usepackage{tikz} % for tikzpicture
\setbeamertemplate{footline}[frame number] 

\mode<presentation>
% \mode<handout>{\setbeamercolor{background canvas}{bg=black!5}}



\title{Random Explanatory variables\footnote{See last slide for copyright information.}}
\subtitle{STA 2101 Fall 2019}
\date{} % To suppress date


\begin{document}

\begin{frame}
  \titlepage
\end{frame}


\begin{frame}
\frametitle{Overview}
\tableofcontents
\end{frame}


\section{Preparation}

\begin{frame}{Preparation: Indicator functions}
\framesubtitle{Conditional expectation and the Law of Total Probability} \pause 

$I_A(x)$ is the \emph{indicator function} for the set $A$. It is defined by
\begin{displaymath}
    I_A(x) = 
\left\{ \begin{array}{ll} % ll means left left
   1 & \mbox{for } x \in A  \\
   0     & \mbox{for } x \notin A 
   \end{array}  \right.   % Need that crazy invisible right period!
\end{displaymath}

\pause 

Also sometimes written $I(x \in A)$

\pause

\begin{eqnarray*}
    E(I_A(X)) &=& \sum_x I_A(x) p(x)  \mbox{, or}\\ \pause
              & & \int_{-\infty}^\infty I_A(x) f(x) \, dx \\ \pause
              &&\\
              &=& P\{ X \in A \}
\end{eqnarray*}
\pause
So the expected value of an indicator is a probability.
\end{frame}


\begin{frame}
\frametitle{Applies to conditional probabilities too} \pause 
%\framesubtitle{} 
{\LARGE
\begin{eqnarray*}
    E(I_A(X)|Y) &=& \sum_x I_A(x) p(x|Y)  \mbox{, or}\\ \pause
              & & \int_{-\infty}^\infty I_A(x) f(x|Y) \, dx \\ \pause
              &&\\
              &=& Pr\{ X \in A|Y\}
\end{eqnarray*}   
} % End size
\pause
So the conditional expected value of an indicator is a \emph{conditional} probability.
\end{frame}

\begin{frame}
\frametitle{Double expectation}
%\framesubtitle{} 
{\LARGE
\begin{displaymath}
    E\left( X \right) = E\left( E[X|Y]\right) = E(g(Y))
\end{displaymath}
} % End size
\end{frame}

\begin{frame}
\frametitle{Showing $E\left( X \right) = E\left( E[X|Y]\right)$}
\framesubtitle{Again note $E\left( E[X|Y]\right)$ is an example of $E(g(Y)$}  \pause
\begin{eqnarray*}
   E\left( E[X|Y]\right) & = & \int  E[X|Y=y] f_y(y) \, dy \\ \pause
   & = & \int  \left(\int x f_{x|y}(x|y) \, dx \right) f_y(y) \, dy \\ \pause
   & = & \int  \left(\int x \frac{f_{x,y}(x,y)}{f_y(y)} \, dx \right) f_y(y) \, dy \\ \pause
   & = & \int  \int x \, f_{x,y}(x,y) \, dx \, dy \\ \pause
   & = & E(X)
\end{eqnarray*} 
\end{frame}


\begin{frame}
\frametitle{Double expectation: $E\left( g(X) \right) = E\left( E[g(X)|Y]\right)$}
%\framesubtitle{} 
\pause 
\begin{displaymath}
    E\left(E[I_A(X)|Y]\right) \pause = E[I_A(X)] \pause = Pr\{ X \in A\} \mbox{, so}
\end{displaymath}
\pause 
\begin{eqnarray*}
    Pr\{ X \in A\} &=&  E\left(E[I_A(X)|Y]\right) \pause \\
              &=& E\left(Pr\{ X \in A|Y\}\right) \pause \\
              &=& \int_{-\infty}^\infty Pr\{ X \in A|Y=y\} f_Y(y) \, dy \mbox{, or} \pause \\
              & & \sum_y Pr\{ X \in A|Y=y\} p_Y(y)
\end{eqnarray*}   

\pause 
This is known as the \emph{Law of Total Probability}
\end{frame}

\section{Random Explanatory Variables}

\begin{frame}
\frametitle{Don't you think it’s strange?}
%\framesubtitle{} 
\begin{itemize}
    \item In the general linear regression model, the $X$ matrix is supposed to be full of fixed constants.  \pause
    \item This is convenient mathematically. \pause Think of $E(\widehat{\boldsymbol{\beta}})$. \pause
    \item But in any non-experimental study, if you selected another sample you'd get different $X$ values\pause, because of random sampling. \pause
    \item So $X$ should be at least partly random variables, not fixed.  \pause
    \item View the usual model as \emph{conditional} on $\mathcal{X}=X$.  \pause
    \item All the probabilities and expected values in the typical regression course are \emph{conditional} probabilities and \emph{conditional} expected values.  \pause
    \item Does this make sense?
\end{itemize}
\end{frame}

\begin{frame}
\frametitle{$\widehat{\boldsymbol{\beta}}$ is (conditionally) unbiased}
%\framesubtitle{} 
{\Large
\begin{displaymath}
    E(\widehat{\boldsymbol{\beta}}|\mathcal{X}=X) = \boldsymbol{\beta}
    \pause \mbox{ for \emph{any} fixed } X
\end{displaymath} 
} % End size

\vspace{5mm} \pause

It's \emph{unconditionally} unbiased too. \pause
 
\vspace{5mm}
{\Large
\begin{displaymath}
    E\{\widehat{\boldsymbol{\beta}}\} \pause = E\{E\{\widehat{\boldsymbol{\beta}}|X\}\} \pause
    = E\{\boldsymbol{\beta}\} \pause = \boldsymbol{\beta}
\end{displaymath} 
} % End size

\end{frame}


\begin{frame}
\frametitle{Perhaps Clearer}
%\framesubtitle{} 

\begin{eqnarray*} 
    E\{\widehat{\boldsymbol{\beta}}\}  \pause
    &=&  E\{E\{\widehat{\boldsymbol{\beta}}|X\}\} 
       \\ \pause
    &=& \int \cdots  \int E\{\widehat{\boldsymbol{\beta}}|\mathcal{X}=X\} \, 
         f(X) \, dX 
        \\ \pause
    &=& \int \cdots  \int \boldsymbol{\beta} \, f(X) \, dX 
         \\ \pause
    &=& \boldsymbol{\beta} \int \cdots  \int f(X)\, dX
        \\ \pause
    &=& \boldsymbol{\beta} \cdot 1 = \boldsymbol{\beta}.
\end{eqnarray*} 
\end{frame}

\begin{frame}
\frametitle{Conditional size $\alpha$ test, Critical value $f_\alpha$} \pause
%\framesubtitle{} 
{\LARGE
\begin{displaymath}
    Pr\{F > f_\alpha | \mathcal{X}=X \} = \alpha
\end{displaymath} \pause
} % End size
% \vspace{3mm}
\begin{eqnarray*}
    Pr\{F > f_\alpha \} &=& \int \cdots  \int Pr\{F > f_\alpha | \mathcal{X}=X \}
                       f(X)\, dX \\ \pause
                   &=& \int \cdots  \int \alpha 
                       f(X)\, dX  \\     \pause
               &=&  \alpha \int \cdots  \int  
                       f(X)\, dX  \\ \pause
                   &=& \alpha  
\end{eqnarray*} 
\end{frame}



\begin{frame}
\frametitle{The moral of the story} \pause
%\framesubtitle{} 
\begin{itemize}
    \item Don't worry. \pause
    \item Even though the explanatory variables are often random, we can apply the usual fixed $X$ model without fear. \pause
    \item Estimators are still unbiased. \pause
    \item Tests have the right Type I error probability. \pause
    \item Similar arguments apply to confidence intervals and prediction intervals. \pause
    \item And it's all distribution-free with respect to $X$.
\end{itemize}
\end{frame}









%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\begin{frame}
\frametitle{Copyright Information}

This slide show was prepared by  \href{http://www.utstat.toronto.edu/~brunner}{Jerry Brunner},
Department of Statistics, University of Toronto. It is licensed under a 
\href{http://creativecommons.org/licenses/by-sa/3.0/deed.en_US}
     {Creative Commons Attribution - ShareAlike 3.0 Unported License}. Use any part of it as you like and share the result freely. The \LaTeX~source code is available from the course website:
\href{http://www.utstat.toronto.edu/~brunner/oldclass/2101f19} {\small\texttt{http://www.utstat.toronto.edu/$^\sim$brunner/oldclass/2101f19}}

\end{frame}

\end{document}