-
Notifications
You must be signed in to change notification settings - Fork 0
/
wip-nw-20210517.tex
executable file
·153 lines (136 loc) · 4.53 KB
/
wip-nw-20210517.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
% Method of the month: The Newey-West estimator and interrupted time series
% (c) 2021 Malcolm Gillies <malcolm.gillies@unsw.edu.au>
% https://github.com/mbg-unsw/neweywest
%
% This work is licensed under a
% Creative Commons Attribution-NonCommercial-ShareAlike 4.0
% International Licence
\documentclass[aspectratio=169,12pt]{beamer} % XXXX fix AR here
\usepackage[latin1]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{textcomp}
\usefonttheme{serif} % need this with Charter font
\usetheme{Berlin} % using default now
\usecolortheme{beaver} % using default now
\usepackage[libertine]{libertine} % not using osf (old-style figures)
\usepackage[scale=0.9]{tgheros} % scale to match libertine
\usepackage[varqu,varl]{inconsolata}
\usepackage[libertine]{newtxmath}
\usepackage{graphicx}
\usepackage{tikz}
\usepackage{tikzpagenodes}
\usepackage{natbib}
\usepackage{gitinfo2}
\renewcommand{\gitMark}{\color{gray}\texttt{\tiny\gitBranch\,@\,\gitAbbrevHash\,\gitAuthorDate}}
\setbeamertemplate{navigation symbols}{} % remove navigation symbols
\setbeamercolor*{item}{fg=darkred}
\title{``Method of the month'': The Newey--West estimator and interrupted time series}
\author{Malcolm Gillies}
\date{20 May 2021}
\usebackgroundtemplate{%
\begin{tikzpicture}[remember picture,overlay]
\node[anchor=south west,scale=1,rotate=90] at ([shift={(0cm,0cm)}]current page marginpar area.south east) {\gitMark};
\end{tikzpicture}%
}
\begin{document}
{
%\usebackgroundtemplate{}
\begin{frame}
\titlepage
\end{frame}
}
\begin{frame}{What is the Newey--West estimator?}
\begin{itemize}
\item Time series data are often \emph{autocorrelated}
\item This messes up inference e.g. 95\% CI too small
\item You can model the correlation structure e.g. ARIMA
\item Or use a Heteroskedasticity- and Autocorrelation-Robust (aka HAC) method
\begin{itemize}
\item e.g. Newey--West (1987) or Newey--West (1994)
\end{itemize}
\end{itemize}
\end{frame}
\begin{frame}{How do I do use it?}
\texttt{library(sandwich)} \\
\texttt{library(lmtest)} \\
\medskip
\texttt{ts.lm <- lm(log(cnt)\textasciitilde 1+step+trend, data=tss)} \\
\texttt{coeftest(ts.lm2, df=Inf, vcov. = NeweyWest(ts.lm2, adjust=TRUE))}
\end{frame}
\begin{frame}{Is anyone else using it?}
\begin{itemize}
\item Review of ITS in healthcare
\citep{hudson_methodology_2019}
\begin{itemize}
\item 2/116 studies published in 2015 used N--W
\end{itemize}
\item Review of ITS studies of public health interventions
\citep{turner_design_2020}
\begin{itemize}
\item 2/200 studies publishd 2013--17 used N--W
\end{itemize}
\end{itemize}
\end{frame}
\usebackgroundtemplate{\includegraphics[width=\paperwidth]{ref/newey-west-cites.PNG}}
\begin{frame}[plain,b]
\begin{flushright}\texttt{https://app.dimensions.ai/}\end{flushright}
\end{frame}
\usebackgroundtemplate{}
\begin{frame}{Empirical comparison 1}
\begin{itemize}
\item Simulation study
\citep{turner_evaluation_2020}
\item Models fit with six methods:
\begin{itemize}
\item OLS, OLS/N--W, GLS/P--W, REML, REML/Satterthwaite, ARIMA
\end{itemize}
\item AR(1) data, N--W \texttt{lag=1}
\item For inference \( T=48 \)
\begin{itemize}
\item ARIMA coverage OK
\item REML similar to ARIMA, REML/Satterthwaite >100\% coverage
\item \textbf{N--W poor coverage}
\end{itemize}
\end{itemize}
\end{frame}
\begin{frame}{Empirical comparison 2}
\begin{itemize}
\item Published ITS studies (\( n=190 \), median \(T=48\))
\citep{turner_comparison_2020}
\item Data re-fit with six methods:
\begin{itemize}
\item OLS, OLS/N--W, GLS/P--W, REML, REML/Satterthwaite, ARIMA
\end{itemize}
\item For inference
\begin{itemize}
\item REML/Satterthwaite most conservative
\item ARIMA 2nd most conservative
\item REML 3rd most conservative
\end{itemize}
\end{itemize}
\end{frame}
\begin{frame}{Bottom line}
\begin{itemize}
\item Don't use N--W (1987) with \texttt{lag=1}
\item N--W (1994) with small sample adjustment might be OK
\item AR(1) with REML estimation sounds good. How do we do it?
\item Post-Newey--West methods tantalizingly close!
\item Maybe stick with ARIMA after all?
\end{itemize}
\end{frame}
\begin{frame}{What do the econometricians say?}
HAR Inference: Recommendations for Practice
\cite{lazarus_har_2018}
\end{frame}
\begin{frame}{Thanks}
\begin{itemize}
\item Andrea
\end{itemize}
\end{frame}
\begin{frame}{References}
% \bibliographystyle{apalike}
% \bibliographystyle{abbrv}
\tiny\bibliography{nw.bib}
\bibliographystyle{abbrvnat}
\end{frame}
\end{document}