Revision

Dr Aishwarya Krishna
 Linear and Multivariate
Regression
Panel regression Stationarity
Dynamic dependence Models AR
Autocorrelation MA
Theories ARMA
Time series analysis Modeling
𝑟𝑡 Forecasting ARIMA

Modelling volatility 𝜎𝑡2 ARCH

GARCH

Non-synchronous trading
High Frequency Data and Algorithmic Trading Bid Ask Spread
Empirical characteristics of data
Modelling

Investments Portfolio optimization

Simple Monte Carlo
Monte Carlo Methods for Option Pricing Antithetic
Estimation
• Point estimate vs Interval Estimate
• Desirable properties of an estimator
• Unbiasedness
• Efficiency
• Consistency
• Linear
• CLT
• Confidence Interval
Regression
• Linearity in variables and parameters
• Why do we need a scatter plot?
• Minimization problem
• Gauss Markov Assumptions
• Coefficient of determination
• Coefficient of correlation
• Multicollinearity
• Heteroskedasticity
• Difference between in F test and t test
Panel data models
• Long and short panels
• Time invariant and Individual invariant regressors
• Overall, between and within variation
• Difference operator
• Pooled models
• Between estimator
• Fixed effects models or within estimator
• Random effects models
Within estimator or Fixed effects
 Random
effects
The RE model assumes that the
individual-specific effects
𝛼𝑖𝑡 distributed independently of the
regressors. We include 𝛼𝑖𝑡 in the
error term.

𝑉𝑎𝑟 𝜈𝑖𝑡 = 𝜎 2𝛼 + 𝜎𝑢2

𝑐𝑜𝑣 𝜈𝑖𝑡 , 𝑣𝑖𝑠 = 𝜎𝛼2
𝜎𝛼2
𝜌𝜈 = 𝑐𝑜𝑟 𝜈𝑖𝑡 , 𝑣𝑖𝑠 = = theta in R
𝜎2𝛼 +𝜎𝑢2
Panel data models
• GLS transformation in random effects
• Estimator model vs true model: consistent estimator
• Breusch Pagan Lagrange Multiplier test
• Hausman test
• Differences between random effects and fixed effects
Time series analysis
• Stationarity in mean, variance, covariance, weak
stationarity and strict stationarity
Identification of the model structure
• Dickey-Fuller Test for Stationarity
1. Identify if the series is stationary
• Test for normality: KS test, Shapiro Wilk test Plot the correlogram (correlogram shows
• Autocorrelation rapid decay if series is stationary)
• Correlogram 2. Remove non-stationarity if any by
• Autocorrelation function differencing/standardization
• Ljung-Box test for serial correlation: null hypothesis of 3. Obtain the AR or MA components/order in
independence in a given time series, residuals are the model
independently distributed. 4. PACF determines the order of the AR model
• AR model: 𝑟𝑡 = 𝛿 + 𝜙1 𝑟𝑡−1 + 𝜔𝑡 : PACF 5. ACF is useful in determining the order of
• MA model: 𝑟𝑡 = 𝛿 + 𝜖𝑡 + 𝜃𝜖𝑡−1 : ACF the MA model
𝑝 𝑞
• ARMA: 𝑟𝑡 = 𝛿 + σ𝑖=1 𝜙𝑖 𝑟𝑡−𝑖 + 𝜖𝑡 + σ𝑖=1 𝜃𝑖 𝜖𝑡−𝑖
• Residual plot to test for normality
Modelling volatility
• What is volatility?
• Why is studying volatility important?
• Characteristics of Volatility
• There exist volatility clusters
• Volatility evolves continuously over time i.e volatility jumps are not frequent
• Volatility does not diverge to infinity. It always has bounds. i.e it is stationary
• Leverage effect: volatility seems to react differently to a big price increase or drop

• ARCH effects checking

• Modelling
• Forecasting
Different models
ARCH EWMA Model GARCH Model
GARCH(m,s)
𝑆𝑡
𝑟𝑡 = log 𝑆𝑡−1
as 𝑟𝑡 = 𝜇𝑡 + • For a log return series For a log return series 𝑟𝑡 =
𝑎𝑡 , 𝑆𝑡
log as 𝑟𝑡 = 𝜇𝑡 + 𝑎𝑡 ,
𝑆𝑡−1
𝑆𝑡
𝑟𝑡 = log as 𝑟𝑡 = 𝜇𝑡 + 𝑎𝑡 ,
𝑆𝑡−1
2 2
2
𝜎𝑡2 = 𝜔 + 𝛼1 𝑎𝑡−1 2 2 𝜎𝑡2 = 𝜔 + 𝛼1 𝑎𝑡−1 + 𝛽1 𝜎𝑛−1
• 𝜎𝑡2 = 𝜆𝜎𝑛−1 + 1 − 𝜆 𝑎𝑛−1 𝑎𝑡 = 𝜖𝑡 × 𝜎𝑡
𝜔 𝜔 • 𝛼1 + 𝛽1 + 𝛾 = 1
𝑉𝐿 = =
1−𝛼1 𝛾 𝜔 𝜔
• 𝑉𝐿 = =
1−𝛼1 −𝛽1 𝛾
• 𝛼1 + 𝛽1 < 1
Understanding R results
• Optimal Parameters: mu, omega, alpha, beta
• Information Criteria
• Ljung-Box Test
• Adjusted Pearson Goodness-of-Fit Test
• sGARCH, TGARCH, GJR-GARCH
Question
Look at the QQ plot below and select the model.
High frequency data
• What is Market microstructure? What is High-frequency financial
data?
• TAQ Database
• Data Characteristics
• Bid Ask spread
• Why is Market microstructure different and difficult?
• Bid Ask Bounce
• Two way classification of price movements
• Duration models
 Different models
Duration model GARCH Model
GARCH(m,s)
𝑥𝑖 = Δ𝑡𝑖 = 𝜓𝑖 𝜖𝑖 𝑎𝑡 = 𝜖𝑡 𝜎𝑡
𝜓𝑖 = 𝐸 𝑥𝑖 𝐹𝑖−1
𝐹𝑖−1 = information set at i − 1 𝑡ℎ trade
𝜎𝑡2 = var(R t |Ft−1 )
𝜖𝑡 = 𝑁(0,1)
𝜖𝑖 ~𝑒𝑥𝑝𝑜𝑛𝑒𝑛𝑡𝑖𝑎𝑙
𝜖𝑖 ~𝑊𝑒𝑖𝑏𝑢𝑙𝑙 2 2
𝜎𝑡2 = 𝜔 + 𝛼1 𝑎𝑡−1 + 𝛽1 𝜎𝑛−1
p
𝜓𝑖 = 𝜔 + Σ𝑖 𝛼𝑖 𝑥𝑖−𝑗 + Σ𝑗 𝛽𝑖 𝜓𝑖−𝑗
𝑞 • 𝛼1 + 𝛽1 + 𝛾 = 1
𝜔 𝜔
EACD(1,1) • 𝑉𝐿 = =
1−𝛼1 −𝛽1 𝛾
𝜓𝑖 = 𝜔 + 𝛼1 𝑥𝑖−1 + 𝛽1 𝜓𝑖−1
conditional expectation of the duration will depend upon past • 𝛼1 + 𝛽1 < 1
durations

Conditional exp duration=f( past durations, past exp duration)

Portfolio Optimization
• Expected Portfolio returns and risk
• VaR and cVaR
• Common portfolios
• Global Minimum Variance (see Markowitz, 1952, 1956, 1991): Based on Variance-Covariance
• Equal Risk Contributed (see Qian, 2005, 2006; Maillard et al., 2010; Qian, 2011): Based on
variance-covariance, marginal risk contributions are equated
• Most Diversified (see Choueifaty and Coignard, 2008; Choueifaty et al., 2011): Based on (i)
correlation matrix and (ii) re-scaling of weights according to assets’ riskiness
• Optimal Tail Dependent: (i) Minimum tail dependent allocation, (ii) Selection of portfolio
constituents from a set of assets
• Comparing the allocations
• Sharpe Ratio
• Efficient Frontier
Option Pricing
• Black Scholes Model
• Using R for Black Scholes Model
• CRR
• CRR model in R
• Connection between the models
• Greeks
• Implied Volatility
Monte Carlo Methods
• Define simulation
• Integration using simulation
• Random variable-> Stochastic Process->Stochastic differential equation-
>Ito Integral->Brownian Motion
• Simulate a price path
• Simulation acceleration techniques
• Monte Carlo price function
• Comparing it with Black Scholes price
• Parallelization
• Advantages of Monte Carlo