Statistics 434: Bullet Points for Day 2 Noise, AR(1), S-Plus, Estimation, and Simulation
We begin with an exploration of the normal noise model, then we look at its simplest alternative, the AR(1) model. This gives us our rst encounter with the notion of
Statistics 434: Bullet Points for Day 22 Comparing Asset Returns in the Context of Risks
Any asset manger, asset class, or investment strategy will be judged on the basis of the historical returns viewed in the context of the risks that were taken but how
Statistics 434: Bullet Points for Day 23 Look Back and Forward Accomplishment and Anticipation
The mission statement for the course has been at the top of the course home page for the whole semester. One may see it at rst, but after a while it surely beco
Statistics 434 Homework No. 2: An Experiment with Ljung-Box
First recall the Ljung-Box Statistic on k lags:
k
LB = T (T + 2)
j=1
j 2 T j
Write an S-Plus function that computes a vector that holds the rst 25 sample autocorrelations of a time series. Hint:
Statistics 434 Homework No. 3: WRDS and Testing for Normality
As preparation, you should skim the material in Zivot and Wang on the creation of time series objects and the use of the timeDate() function, but the e-Handout WRDStoFinMet.txt will have most o
Statistics 434 Homework No. 4 ARMA, ACF, PACF, and a Betting Simulation
Reading Review the material in Zivot and Wang on arima.sim, arima.mle, and arima.forecast. Also review the information on these functions using the S-Plus help(). Simulation and Estim
Statistics 434 Homework No. 5: Kelly Betting on AR(1)
This homework is more open-ended than those we have done before. It provides plenty of room for you to exercise your good sense and to show that you can conduct a wise exploration of the ideas that hav
Statistics 434: Homework No. 6 Stationarity, Extremes, and Opportunities
Part 1: Checking Out the Unit Root Tests Obtain 500 days of price and return data for a stock of your choosing. Use the unitroot() function to nd the p-values for the augmented Dicke
Statistics 434: Homework No. 8
Data Get four years (or a little less) of daily returns from CRSP for one of your favorite rms, or perhaps of some ragged mutual fund, or some interesting ETF. For this assignment youll need a time series of length 1000 or s
Statistics 434: Bullet Points for Day 1 Getting Started Big Picture, S-Plus, Look at a Model
The Bullet Point Day Plans serve to organize the class time and to provide you with quick reviews of what topics were covered. These outline are but faint shadows
Statistics 434: Bullet Points for Day 4 Autocorrelation in Theory, Practice, and Tests
The rst question that one must ask of a stationary time series is simply Do I have any reason to believe that there is any dependence among these values, or do they beh
Statistics 434: Bullet Points for Day 5 Autocorrelation Tests Especially the Ljung-Box Test
When is it feasible to treat a series of returns as if they are independent? In other words, when do returns behave like noise? This is one of the most basic quest
Statistics 434: Bullet Points for Day 6 WRDS, CRSP, Real Asset Returns, and Normality Assumptions
The main task today is rock-bottom practical: How does one access the CRSP data via WRDS? We also ask "When is it feasible to treat a series of returns as if
Statistics 434 : Bullet Points for Day 9 Simulating and Fitting ARIMA(p,d,q) Models
The main formal task is to pick up the computation tools for studying the ARIMA(p,d,q) models. In particular, we consider the tools for simulation and for tting using the
Statistics 434: Bullet Points for Day 10 Bet Sizing and Long-term Wealth Growth Rates
Today we take a little side tour from our investigation of the ARIMA(p,d,q) model to look at what the general question "Facing the possibility of a favorable bet, what f
Statistics 434: Bullet Points for Day 21 Co-integration and Statistical Arbitrage
When one reasons about asset prices, there is a recurring tension between "trend following" and "mean reversion." So far, almost all of the tools that we have considered in
Bullet Points for Day 20 Rolling Statistics and Momentum Strategies
It never makes sense to ignore data, and we always want to use all the data that is available to us at the time we must act. This means we need to compute many statistics in a "rolling" f
Bullet Points for Day 18 After GARCH? In Comes the GARCH ZOO!
We begin with a celebration of the model with that is AR(1) in "mean" and GARCH(1,1) in "error." It's claim to fame is that it is our first genuinely feasible candidate for an asset return time
Statistics 434: Bullet Points for Day 3
AR(1) Estimation Point Estimates and Their Distributions
The AR(1) model will serve as our model for a model. That is, what we
can see and say about AR(1) provides an outline of what we would hope to see
or say abou
Statistics 434: Bullet Points for Day 7 AR(p) with p 2
We now consider the autoregressive models with p 2, beginning with a look at the most useful of these the modest AR(2) model that was made famous by Yules study of sun spot data. We need to understand
Statistics 434: Bullet Points for Day 8
ARIMA(p,d,q) In Full
After considering the stylized facts suggested by HW3, we complete our
description of the most widely used class of univariate time series models, the
ARIMA(p,d,q) models. Our main task is to un
Bullet Points for Day 19
Comparing GARCH Family Members
The large number of models in the GARCH family may give one an uneasy
feeling. With so many alternatives, how can one make a reasonable choice?
Some criteria for choosing among models are tness for u
Statistics 434 Homework No. 1
Experience with S functions and Simulation Write a function getrhohat() which creates a realization of , the sample autocorrelation for the series cfw_yt which is a simulated realization of length T = 100 from a stationary A
Statistics 434: Homework No. 7
Exploration of Style Momentum
In the language of mutual funds, style has come to refer to the nine boxes
into which Morningstar jams all funds. These styles are given by the cross table
of (small, medium, large) capitalizati
Statistics 434: Homework No. 9
Reading Read the S-Plus Help les for all of the new S-Plus functions that we have covered and explore the help les for the related functions that are given at the bottom of the help les. Data Get four years (or a little less
Statistics 434: Bullet Points A Pre-Thanksgiving Interlude
In the Wednesday afternoon before Thanksgiving, attendance typically runs under 60%, so the most practical to use of this time is to cover some special topics that one can miss without creating a
Statistics 434: Bullet Points for Day 12 Martingales: The Most Important Stochastic Processes
At the rst level, martingales are simple mathematical objects that help us understand fair games, including the impossibility of eective gambling systems. Its no
Statistics 434: Bullet Points for Day 13 Stationarity and Unit Root Tests
If a time series is not stationary, or cannot be transformed to be stationary say by taking dierences, then we are pretty close to stuck. Weve lost our major hold on learning from t
Bullet Points for Day 14 Value at Risk, Extreme Values, Risk-Adjusted Returns
When the time comes to quantifying risk, you almost have to turn to the alleged Yogi quote: In theory, theory and practice are equivalent. In practice, they are not. The somber
Statistics 434: Bullet Points for Day 15 Time Series Regression, CAPM, and the Three Factor Model
For us, time series regressions are typied by the Capital Asset Pricing Model (CAPM) and the Fama French Three Factor Model (FF3FM). Well also add to your to