CFA二级培训项目 Quantitative Methods CFA二级课程框架 10 Study Session 1-2 Ethics & Professional Standards Study Session 3 Quantitative Analysis 5-10 Study Session 4 Economics 5-10 Study Session 5-7 Financial Reporting and Analysis 15-25 Study Session 8-9 Corporate Finance 5-15 Study Session 18 Portfolio Management and Wealth Planning 5-15 Study Session 10-12 Equity Investment 20-30 Study Session 14-15 Fixed Income 5-15 Study Session 16-17 Derivatives 5-15 Study Session 13 Alternative Investments 5-15 Total: 100 2-79 Summary of Readings and Framework SS 3 R9 Correlation and regression R10 Multiple regression and issues in regression analysis R11 Time-series analysis R12 Excerpt from ’’Probabilistic Approaches: Scenario Analysis, Decision Trees, and Simulation’’ 3-79 Framework Correlation and Regression 1. Scatter Plots 2. Covariance and Correlation 3. Interpretations of Correlation Coefficients 4. Significance Test of the Correlation 5. Limitations to Correlation Analysis 6. The Basics of Simple Linear Regression 7. Interpretation of regression coefficients 8. Standard Error of Estimate & Coefficient of Determination (R2) 9. Analysis of Variance (ANOVA) 10. Regression coefficient confidence interval 11. Hypothesis Testing about the Regression Coefficient 12. Predicted Value of the Dependent Variable 13. Limitations of Regression Analysis 4-79 Scatter Plots A scatter plots is a graph that shows the relationship between the observations for two data series in two dimensions. 5-79 Covariance and Correlation Covariance: Covariance measures how one random variable moves with another random variable. ----It captures the linear relationship. n Cov( X , Y ) ( X i X )(Yi Y ) /( n 1) i 1 Covariance ranges from negative infinity to positive infinity Correlation: Cov( X , Y ) r sx s y Correlation measures the linear relationship between two random variables Correlation has no units, ranges from –1 to +1 6-79 Interpretations of Correlation Coefficients The correlation coefficient is a measure of linear association. It is a simple number with no unit of measurement attached, so the correlation coefficient is much easier to explain than the covariance. 7-79 Correlation coefficient Interpretation r = +1 perfect positive correlation 0 < r < +1 positive linear correlation r=0 no linear correlation −1 < r < 0 negative linear correlation r = −1 perfect negative correlation Interpretations of Correlation Coefficients 8-79 Significance Test of the Correlation Test whether the correlation between the population of two variables is equal to zero. H0: ρ=0 t-test t = r n-2 1-r 2 , df = n-2 Two-tailed test Decision rule: reject H0 if +t critical <t, or t<- t critical 9-79 Significance Test of the Correlation Example: An analyst is interested in predicting annual sales for XYZ Company, a maker of paper products. The following table reports a regression of the annual sales for XYZ against paper product industry sales. The correlation between company and industry sales is 0.9757. The regression was based on five observations. Coefficient Standard error of the coefficient Intercept -94.88 32.97 Slope (industry sales) 0.2796 0.0363 t= r n2 1 r2 0.9757 3 7.72 1 0.952 From the t-table, we find that with df = 3 and 95% significance, the two-tailed critical t-values are ±3.182 (recall that for the t-test the degrees of freedom = n 2). Because the computed t is greater than +3.182, the correlation coefficient is significantly different from zero. 10-79 Limitations to Correlation Analysis Outliers (异常值) Outliers represent a few extreme values for sample observations. Relative to the rest of the sample data, the value of an outlier may be extraordinarily large or small. Outlier can result in apparent statistical evidence that a significant relationship exists when, in fact, there is none, or that there is no relationship when, in fact, there is a relationship. 11-79 Limitations to Correlation Analysis Spurious correlation (假相关) Spurious correlation refers to the appearance of a causal linear relationship when, in fact, there is no relation. Certain data items may be highly correlated purely by chance. That is to say, there is no economic explanation for the relationship, which would be considered a spurious correlation. 12-79 Limitations to Correlation Analysis Nonlinear relationships Correlation only measures the linear relationship between two variables, so it dose not capture strong nonlinear relationships between variables. For example, two variables could have a nonlinear relationship such as Y= (1-X) 3 and the correlation coefficient would be close to zero, which is a limitation of correlation analysis. 13-79 The Basics of Simple Linear Regression Linear regression allows you to use one variable to make predictions about another, test hypotheses about the relation between two variables, and quantify the strength of the relationship between the two variables. Linear regression assumes a linear relation between the dependent and the independent variables. The dependent variable (y) is the variable whose variation is explained by the independent variable. The dependent variable is also refer to as the explained variable, the endogenous variable,or the predicted variable. The independent variable (x) is the variable whose variation is used to explain the variation of the dependent variable. The independent variable is also refer to as the explanatory variable, the exogenous variable, or the predicting variable. 14-79 The Basics of Simple Linear Regression The simple linear regression model Yi b0 b1 X i i , i 1,..., n Where, Yi = ith observation of the dependent variable, Y Xi = ith observation of the independent variable, X b0 = regression intercept term b1 = regression slope coefficient εi= the residual for the ith observation (also referred to as the disturbance term or error term) 15-79 The Basics of Simple Linear Regression The assumptions of the linear regression A linear relationship exists between X and Y X is not random. (even if X is random, we can still rely on the results of regression models given the crucial assumption that X is uncorrelated with the error term). The expected value of the error term is zero (i.e., E(εi)=0 ) The variance of the error term is constant (i.e., the error terms are homoskedastic) The error term is uncorrelated across observations (i.e., E(εiεj)=0 for all i≠j) The error term is normally distributed. 16-79 Interpretation of regression coefficients Interpretation of regression coefficients The estimated intercept coefficient ( b̂0 ) is interpreted as the value of Y when X is equal to zero. The estimated slope coefficient ( b̂1 ) defines the sensitivity of Y to a change in X .The estimated slope coefficient ( b̂1 ) equals covariance divided by variance of X. n Cov( X , Y ) b1 Var ( X ) (X i 1 n 17-79 X )(Yi Y ) 2 ( X X ) i i 1 b0 Y b1 X i Interpretation of regression coefficients Example An estimated slope coefficient of 2 would indicate that the dependent variable will change two units for every 1 unit change in the independent variable. The intercept term of 2% can be interpreted to mean that the independent variable is zero, the dependent variable is 2%. 18-79 An example: calculate a regression coefficient The individual observations on countries' annual average money supply growth from 1970-2001 are denoted Xi, and individual observations on countries' annual average inflation rate from 1970-2001 are denoted Y. Cross-Product Squared Deviations Squared Deviations (Xi - X )( Yi- Y ) (Xi - X )2 (Yi - Y)2 0.0676 0.000169 0.000534 0.000053 0.0915 0.0519 0.000017 0.000004 0.000071 New Zealand 0.1060 0.0815 0.000265 0.000156 0.000449 Switzerland 0.0575 0.0339 0.000950 0.001296 0.000697 United Kingdom 0.1258 0.0758 0.000501 0.001043 0.000240 United States 0.0634 0.0509 0.000283 0.000906 0.000088 Sum 0.5608 0.3616 0.002185 0.003939 0.001598 Average 0.0935 0.0603 Variance 0.000788 0.000320 Standard deviation 0.028071 0.017889 Country Money Supply Growth Rate Xi Inflation Rate Yi Australia 0.1166 Canada Covariance 19-79 0.000437 An answer: calculate a regression coefficient n bˆ 1 = Y -Y X -X i i i=1 n X -X 2 0.000437 = = 0.5545, and 0.000788 i i=1 bˆ 0 Y bˆ 1 X 0.0603-0.5545(0.0935) 0.0084 Y 0.0084 0.5545 X 20-79 Standard Error of Estimate & Coefficient of Determination (R2) Standard Error of Estimate (SEE) measures the degree of variability of the actual Y-values relative to the estimated Y-values from a regression equation. SEE will be low (relative to total variability) if the relationship is very strong and high if the relationship is weak. The SEE gauges the “fit” of the regression line. The smaller the standard error, the better the fit. The SEE is the standard deviation of the error terms in the regression. 21-79 Standard Error of Estimate & Coefficient Determination (R2) The Coefficient Determination (R2) is defined as the percentage of the total variation in the dependent variable explained by the independent variable. Example: R2 of 0.63 indicates that the variation of the independent variable explains 63% of the variation in the dependent variable. 22-79 ANOVA Table ANOVA Table df SS MSS Regression k=1 RSS MSR=RSS/k Error n-2 SSE MSE=SSE/(n-2) Total n-1 SST - Standard error of estimate SEE SSE MSE n2 Coefficient of determination (R²) 23-79 SSR SSE 1 SST SST explained variation unexplained variation =1total variation total variation R2 For simple linear regression, R²is equal to the squared correlation coefficient (i.e., R²= r²) Example: An analyst ran a regression and got the following result: Coefficient t-statistic p-value Intercept -0.5 -0.91 0.18 Slope -2.5 10.00 <0.001 df SS MSS Regression 1 7000 ? Error ? 3000 ? Total 41 ? - Fill in the blanks of the ANOVA Table. What is the standard error of estimate? What is the result of the slope coefficient significance test? What is the result of the sample correlation? What is the 95% confidence interval of the slope coefficient? 24-79 Regression coefficient confidence interval Regression coefficient confidence interval bˆ1 t c sbˆ 查表所得 1 If the confidence interval at the desired level of significance dose not include zero, the null is rejected, and the coefficient is said to be statistically different from zero. sb̂ 1 is the standard error of the regression coefficient. As SEE rises, sb̂ 1 also increases, and the confidence interval widens because SEE measures the variability of the data about the regression line, and the more variable the data, the less confidence there is in the regression model to estimate a coefficient. 25-79 Hypothesis Testing about the Regression Coefficient Significance test for a regression coefficient H0: b1=The hypothesized value bˆ1 b1 t sbˆ df=n-2 1 Decision rule: reject H0 if +t critical <t, or t<- t critical Rejection of the null means that the slope coefficient is different from the hypothesized value of b1 26-79 Predicted Value of the Dependent Variable Predicted values are values of the dependent variable based on the estimated regression coefficients and a prediction about the value of the independent variable. Point estimate Yˆ bˆ0 bˆ1 X ' Confidence interval estimate Yˆ t c s f t c = the critical t-value with df=n−2 s f = the standard error of the forecast 1 ( X ' X )2 1 ( X ' X )2 s f SEE 1 SEE 1 2 n (n 1) s X n ( X i X )2 27-79 Limitations of Regression Analysis Regression relations change over time This means that the estimation equation based on data from a specific time period may not be relevant for forecasts or predictions in another time period. This is referred to as parameter instability. The usefulness will be limited if others are also aware of and act on the relationship. Regression assumptions are violated 28-79 For example, the regression assumptions are violated if the data is heteroskedastic (non-constant variance of the error terms) or exhibits autocorrelation (error terms are not independent). Summary of Readings and Framework SS 3 R9 Correlation and regression R10 Multiple regression and issues in regression analysis R11 Time-series analysis R12 Excerpt from ’’Probabilistic Approaches: Scenario Analysis, Decision Trees, and Simulation’’ 29-79 Framework Multiple Regression 1. The Basics of Multiple Regression 2. Interpreting the Multiple Regression Results 3. Hypothesis Testing about the Regression Coefficient 4. Regression Coefficient F-test 5. Coefficient of Determination (R2) 6. Analysis of Variance (ANOVA) 7. Dummy variables 8. Multiple Regression Assumptions 9. Multiple Regression Assumption Violations 10. Model Misspecification 11. Qualitative Dependent Variables 30-79 The Basics of Multiple Regression Multiple regression is regression analysis with more than one independent variable The multiple linear regression model Yi b0 b1 X 1i b2 X 2i bk X ki i Xij = ith observation of the jth independent variable N = number of observation K = number of independent variables Predicted value of the dependent variable Yˆ bˆ bˆ Xˆ bˆ Xˆ bˆ Xˆ 0 31-79 1 1 2 2 k k Interpreting the Multiple Regression Results The intercept term is the value of the dependent variable when the independent variables are all equal to zero. Each slope coefficient is the estimated change in the dependent variable for a one unit change in that independent variable, holding the other independent variables constant. That’s why the slope coefficients in a multiple regression are sometimes called partial slope coefficient. 32-79 Hypothesis Testing about the Regression Coefficient Significance test for a regression coefficient H0: bj=0 t bˆ j sbˆ df=n-k-1 j p-value: the smallest significance level for which the null hypothesis can be rejected Reject H0 if p-value<α Fail to reject H0 if p-value>α Regression coefficient confidence interval ˆ b j tc sbˆ 33-79 j Estimated regression coefficient ±(critical t-value) (coefficient standard error) Regression Coefficient F-test An F-statistic assesses how well the set of independent variables, as a group, explains the variation in the dependent variable. An F-test is used to test whether at least one slope coefficient is significantly different from zero H0: b1= b2= b3= … = bk=0 Ha: at least one bj≠0 (j = 1 to k) F-statistic 34-79 MSR F MSE SSE SSR k (n k 1) Regression Coefficient F-test Decision rule reject H0 : if F (test-statistic) > F c (critical value) Rejection of the null hypothesis at a stated level of significance indicates that at least one of the coefficients is significantly different than zero, which is interpreted to mean that at least one of the independent variables in the regression model makes a significant contribution on the explanation of the dependent variable. The F-test here is always a one-tailed test The test assesses the effectiveness of the model as a whole in explaining the dependent variable 35-79 Coefficient of Determination (R2) Interpretation The percentage of variation in the dependent variable that is collectively explained by all of the independent variables. For example, an R2 of 0.63 indicates that the model, as a whole, explains 63% of the variation in the dependent variable. Adjusted R2 R2 by itself may not be a reliable measure of the explanatory power of the multiple regression model. This is because R2 almost always increases as variables are added to the model, even if the marginal contribution of the new variables is not statistically significant. 36-79 Analysis of Variance (ANOVA) ANOVA Table d.f. SS MSS Regression k RSS MSR=RSS/k Error n-k-1 SSE MSE=SSE/(n-k-1) Total n-1 SST - Standard error of estimate SSE MSE n k 1 SEE Coefficient of determination (R²) RSS SSE R2 1 SST SST n 1 2 2 adjusted R 1 1 R n k 1 adjusted R²≤ R² adjusted R²may be less than zero 37-79 1 adjusted R 2 1 R2 n 1 n k 1 Dummy variables To use qualitative variables as independent variables in a regression The qualitative variable can only take on two values, 0 and 1 If we want to distinguish between n categories, we need n−1 dummy variables 38-79 Dummy variables Interpreting the coefficients Example: EPSt = b0 + b1Q1t + b2Q2t + b3Q3t + t EPSt = a quarterly observation of earnings per share Q1t =1 if period t is the first quarter, Q1t =0 otherwise Q2t =1 if period t is the second quarter, Q2t =0 otherwise Q3t =1 if period t is the third quarter, Q3t =0 otherwise 39-79 The intercept term, represents the average value of EPS for the fourth quarter. The slope coefficient on each dummy variable estimates the difference in earnings per share (on average) between the respective quarter (i.e., quarter 1, 2, or 3) and the omitted quarter (the fourth quarter in this case). y x1 x2 x3 EPSt Q1 Q2 Q3 EPS09Q4 0 0 0 EPS09Q3 0 0 1 EPS09Q2 0 1 0 EPS09Q1 1 0 0 EPS08Q4 0 0 0 EPS08Q3 0 0 1 EPS08Q2 0 1 0 EPS08Q1 1 0 0 … … … … Unbiased and consistent estimator An unbiased estimator is one for which the expected value of the estimator is equal to the parameter you are trying to estimate. If not, called as unreliable. A consistent estimator is one for which the accuracy of the parameter estimate increases as the sample size increases. 40-79 Multiple Regression Assumptions The assumptions of the multiple linear regression 41-79 A linear relationship exists between the dependent and independent variables The independent variables are not random ( OR X is not correlated with error terms). There is no exact linear relation between any two or more independent variables The expected value of the error term is zero (i.e., E(εi)=0 ) The variance of the error term is constant (i.e., the error terms are homoskedastic) The error term is uncorrelated across observations (i.e., E(εiεj)=0 for all i≠j) The error term is normally distributed Multiple Regression Assumption Violations Heteroskedasticity 异方差 Heteroskedasticity refers to the situation that the variance of the error term is not constant (i.e., the error terms are not homoskedastic) Unconditional heteroskedasticity occurs when the heteroskedasticity is not related to the level of the independent variables, which means that it dose not systematically increase or decrease with the change in the value of the independent variables. It usually causes no major problems with the regression. Conditional heteroskedasticity is heteroskedasticity, that is, variance of error term is related to the level of the independent variables. Conditional heteroskedasticity dose create significant problems for statistical inference. 42-79 Multiple Regression Assumption Violations Effect of Heteroskedasticity on Regression Analysis Not affect the consistency of regression parameter estimators Consistency: the larger the number of sample, the lower probability of error. ˆ The coefficient estimates (the b j ) are not affected. The standard errors are usually unreliable estimates. If the standard errors are too small, but the coefficient estimates themselves are not affected, the t-statistics will be too large and the null hypothesis of no statistical significance is rejected too often (一类错误). The opposite will be true if the standard errors are too large. (二类错误) 43-79 The F-test is also unreliable. Multiple Regression Assumption Violations Detecting Heteroskedasticity Two methods to detect heteroskedasticity (1) residual scatter plots (residual vs. independent variable) (2) the Breusch-Pagen χ² test H0: No heteroskedasticity BP = n×Rresidual² , df=k one-tailed test 注意:以误差项 squred residuals和X做回归,是此回归的决定系数 Decision rule: BP test statistic should be small (χ²分布表) Correcting heteroskedasticity robust standard errors (also called White-corrected standard errors) generalized least squares 44-79 CFA考试不要求了解如何修正方程,只要 知道如果有异方差问题,用robust standard error计算t-statistics Multiple Regression Assumption Violations Serial correlation (autocorrelation) 序列相关,自相关 Serial correlation (autocorrelation) refers to the situation that the error terms are correlated with one another Serial correlation is often found in time series data Positive serial correlation exists when a positive regression error in one time period increases the probability of observing regression error for the next time period. Negative serial correlation occurs when a positive error in one period increases the probability of observing a negative error in the next period. 45-79 Multiple Regression Assumption Violations Effect of Serial correlation on Regression Analysis Positive serial correlation → Type I error & F-test unreliable Not affect the consistency of estimated regression coefficients. Because of the tendency of the data to cluster together from observation to observation, positive serial correlation typically results in coefficient standard errors that are too small, which will cause the computed t-statistics to be larger. Positive serial correlation is much more common in economic and financial data, so we focus our attention on its effects. Negative serial correlation → Type II error (考试不做要求) 46-79 Because of the tendency of the data to diverge from observation to observation, negative serial correlation typically causes the standard errors that are too large, which leads to the computed t-statistics too small. Multiple Regression Assumption Violations Detecting Serial correlation Two methods to detect serial correlation (1) residual scatter plots (2) the Durbin-Watson test H0: No serial correlation DW ≈ 2×(1−r) Decision rule 残差的 correlation 思考:当r=0, 1, -1时候 Reject H0, Reject H0, conclude conclude positive serial Inconclusive Do not Inconclusive negative serial reject H0 correlation correlation 0 d1 dU 4-d1 4 4-dU 47-79 Durbin-Watson test H0: No positive serial correlation DW ≈ 2×(1−r) 48-79 Decision rule Reject H0, conclude positive serial correlation Inconclusive Fail to reject null hypothesis of no positive serial correlation 0 d1 dU Multiple Regression Assumption Violations Methods to Correct Serial correlation adjusting the coefficient standard errors (e.g., Hansen method): the Hansen method also corrects for conditional heteroskedaticity. The White-corrected standard errors are preferred if only heteroskedasticity is a problem. 49-79 Improve the specification of the model: The best way to do this is to explicitly incorporate the time-series nature of the data (e.g., include a seasonal term). Multiple Regression Assumption Violations Multicollinearity 多重共线性 Multicollinearity refers to the situation that two or more independent variables are highly correlated with each other In practice, multicollinearity is often a matter of degree rather than of absence or presence. Two methods to detect multicollinearity (1) t-tests indicate that none of the individual coefficients is significantly different than zero, while the F-test indicates overall significance and the R²is high (2) the absolute value of the sample correlation between any two independent variables is greater than 0.7 (i.e., ︱r︱>0.7) 50-79 Methods to correct multicollinearity: omit one or more of the correlated independent variables Multiple Regression Assumption Violations Summary of assumption violations Assumption violation Impact Detection Conditional Heteroskedasticity Type I /II error ① Residual scatter plots ② Breusch-Pagen χ²-test (BP = n×R²) Positive serial correlation Type I error ① Residual scatter plots ② Durbin-Watson test (DW≈2×(1−r)) Multicollinearity 51-79 Type II error ① t-tests: fail to reject H0; (high R2 and F-test: reject H0; R²is high low t-statistics) ② High correlation among independent variables Model Misspecification There are three broad categories of model misspecification, or ways in which the regression model can be specified incorrectly, each with several subcategories: 1. The functional form can be misspecified. Important variables are omitted. Variables should be transformed. Data is improperly pooled. 2. Explanatory variables are correlated with the error term in time series models. A lagged dependent variable is used as an independent variable. A function of the dependent variable is used as an independent variable ("forecasting the past"). Independent variables are measured with error. 3. Other time-series misspecifications that result in nonstationarity. Effects of the model misspecification: regression coefficients are biased and/or inconsistent 52-79 Qualitative Dependent Variables Qualitative dependent variable is a dummy variable that takes on a value of either zero or one Probit and logit model: Application of these models results in estimates of the probability that the event occurs (e.g., probability of default). A probit model based on the normal distribution, while a logit model is based on the logistic distribution Both models must be estimated using maximum likelihood methods(极 大似然估计) These coefficients relate the independent variables to the likelihood of an event occurring, such as a merger, bankruptcy, or default. Discriminant models yields a linear function, similar to a regression equation, which can then be used to create an overall score, or ranking, for an observation. Based on the score, an observation can be classified into the bankrupt or not bankrupt category. 53-79 Summary of Readings and Framework SS 3 R9 Correlation and regression R10 Multiple regression and issues in regression analysis R11 Time-series analysis R12 Excerpt from ’’Probabilistic Approaches: Scenario Analysis, Decision Trees, and Simulation’’ 54-79 Framework Time-Series Analysis 1. Trend Models 2. Autoregressive Models (AR) 3. Random Walks 4. Autoregressive Conditional Heteroskedasticity (ARCH) 5. Regression with More Than One Time Series 6. Steps in Time-Series Forecasting 55-79 Trend Models Linear trend model yt=b0+b1t+εt Same as linear regression, except for that the independent variable is time t (t=1, 2, 3, ……) yt t 56-79 Trend Models Log-linear trend model yt=e(b0+b1t) Ln(yt ) =b0+b1t+εt Model the natural log of the series using a linear trend Use the Durbin Watson statistic to detect autocorrelation 57-79 Trend Models Factors that Determine Which Model is Best A linear trend model may be appropriate if the data points appear to be equally distributed above and below the regression line (inflation rate data). A log-linear model may be more appropriate if the data plots with a non-linear (curved) shape, then the residuals from a linear trend model will be persistently positive or negative for a period of time (stock indices and stock prices). Limitations of Trend Model Usually the time series data exhibit serial correlation, which means that the model is not appropriate for the time series, causing inconsistent b0 and b1 The mean and variance of the time series changes over time. 58-79 Autoregressive Models (AR) An autoregressive model uses past values of dependent variables as independent variables AR(p) model xt b0 b1 xt-1 b2 xt-2 ... bp xt p t 59-79 AR (p): AR model of order p (p indicates the number of lagged values that the autoregressive model will include). For example, a model with two lags is referred to as a second-order autoregressive model or an AR (2) model. Autoregressive Models (AR) Forecasting With an Autoregressive Model Chain rule of forecasting A one-period-ahead forecast for an AR (1) model is determined in the following manner: xt 1 b0 b1 xt Likewise, a two-step-ahead forecast for an AR (1) model is calculated as: xt 2 b0 b1 xt 1 60-79 Autoregressive Models (AR) Forecasting With an Autoregressive Model, we should prove: No autocorrelation Covariance-stationary series No Conditional Heteroskedasticity 61-79 Autoregressive Models (AR) Detecting autocorrelation in an AR model Compute the autocorrelations of the residual t-tests to see whether the residual autocorrelations differ significantly from 0, t statistics , t t k 1/ n n is the number of observations in the time series. 62-79 If the residual autocorrelations differ significantly from 0, the model is not correctly specified, so we may need to modify it (e.g. seasonality) Correction: add lagged values Autoregressive Models (AR) Covariance-stationary series Statistical inference based on OLS estimates for a lagged time series model assumes that the time series is covariance stationary Three conditions for covariance stationary Constant and finite expected value of the time series Constant and finite variance of the time series Constant and finite covariance with leading or lagged values Stationary in the past does not guarantee stationary in the future All covariance-stationary time series have a finite mean-reverting level. 63-79 Autoregressive Models (AR) Mean reversion A time series exhibits mean reversion if it has a tendency to move towards its mean b0 For an AR(1) model, the mean reverting level is: xt = (1 b1 ) If xt b0 (1 b1 ) than x t, and if higher than x t 64-79 the model predicts that x t+1 will be lower xt b0 (1 b1 ) the model predicts that x t+1 will be Autoregressive model 如果没有 mean reverting level说明follow random walk. Autoregressive Models (AR) Instability of regression coefficients Financial and economic relationships are dynamic Models estimated with shorter time series are usually more stable than those with longer time series So we need to check Covariance stationary 65-79 Compare forecasting power with RMSE Comparing forecasting model performance In-sample forecasts are within the range of data (i.e., time period) used to estimate the model, which for a time series is known as the sample or test period. Root mean squared error (RMSE): the model with the smallest RMSE is most accurate for out-of-sample Out-of-sample forecasts are made outside. In other words, we compare how accurate a model is in forecasting the y variable value for a time period outside the period used to develop the model. 66-79 Random Walks Random walk A special AR(1) model with b0=0 and b1=1 Simple random walk: xt =xt-1+εt The best forecast of xt is xt-1 Random walk with a drift xt=b0+b1 xt-1+εt b0≠0, b1=1 67-79 The time series is expected to increase/decrease by a constant amount Random Walks Covariance stationary A random walk has an undefined mean reverting level A time series must have a finite mean reverting level to be covariance stationary A random walk, with or without a drift, is not covariance stationary The time series is said to have a unit root if the lag coefficient is equal to one Dickey and Fuller test 68-79 Random Walks The unit root test of nonstationarity The time series is said to have a unit root if the lag coefficient is equal to one A common t-test of the hypothesis that b1=1 is invalid to test the unit root Dickey-Fuller test (DF test) to test the unit root Start with an AR(1) model xt=b0+b1 xt-1+εt Subtract xt-1 from both sides xt-xt-1 =b0+(b1 –1) xt-1+εt xt-xt-1 =b0+g xt-1+εt H0: g=0 (has a unit root and is nonstationary) Ha: g<0 (does not have a unit root and is stationary) Calculate conventional t-statistic and use revised t-table If we can reject the null, the time series does not have a unit root and is stationary 69-79 Random Walks – if a time series appears to have a unit root If a time series appears to have a unit root, how should we model it ??? One method that is often successful is to first-difference the time series (as discussed previously) and try to model the firstdifferenced series as an autoregressive time series First differencing e.g. 2, 5, 10, 17, ? , 37 Define yt as yt = xt - xt-1 This is an AR(1) model yt = b0 + b1 yt-1 +εt , where b0=b1=0 The first-differenced variable yt is covariance stationary 70-79 Autoregressive Models (AR) Seasonality –a special question Time series shows regular patterns of movement within the year The seasonal autocorrelation of the residual will differ significantly from 0 We should uses a seasonal lag in an AR model For example: xt=b0+b1 xt-1+ b2 xt-4+εt 71-79 Autoregressive Conditional Heteroskedasticity (ARCH) Heteroskedasticity refers to the situation that the variance of the error term is not constant Test whether a time series is ARCH(1) 多元回归中用BP test t2 a0 a1 t21 ut If the coefficient a1 is significantly different from 0, the time series is ARCH(1) Generalized least squares must be used to develop a predictive model Use the ARCH model to predict the variance of the residuals in following periods 72-79 Regression with More Than One Time Series In linear regression, if any time series contains a unit root, OLS may be invalid Use DF tests for each of the time series to detect unit root, we will have 3 possible scenarios None of the time series has a unit root: we can use multiple regression At least one time series has a unit root while at least one time series does not: we cannot use multiple regression Each time series has a unit root: we need to establish whether the time series are cointegrated. If conintegrated, can estimate the long-term relation between the two series (but may not be the best model of the short-term relationship between the two series). 73-79 Regression with More Than One Time Series Use the Dickey-Fuller Engle-Granger test (DF-EG test) to test the cointegration H0: no cointegration If we cannot reject the null, we cannot use multiple regression If we can reject the null, we can use multiple regression 74-79 Ha: cointegration Summary of Readings and Framework SS 3 R9 Correlation and regression R10 Multiple regression and issues in regression analysis R11 Time-series analysis R12 Excerpt from“Probabilistic Approaches: Scenario Analysis, Decision Trees, and Simulation” 75-79 Simulation Steps in Simulation Determine “probabilistic” variables Define probability distributions for these variables Historical data Cross sectional data Statistical distribution and parameters Check for correlation across variables When two variables are strong correlated, one solution is to pick only one of the two inputs; the other is to build the correlation explicitly into the simulation. 76-79 Run the simulation Simulation Advantage of using simulation in decision making Better input estimation A distribution for expected value rather than a point estimate Simulations with Constraints Book value constraints Regulatory capital restrictions Financial service firms Negative book value for equity Earnings and cash flow constraints Either internally or externally imposed Market value constraints Model the effect of distress on expected cash flows and discount rates 77-79 Simulation Issues in using simulation GIGO Real data may not fit distributions Non-stationary distributions Changing correlation across inputs 78-79 Comparing the Approaches Choose scenario analysis, decision trees, or simulations Selective versus full risk analysis Type of risk Discrete risk vs. Continuous risk Concurrent risk vs. Sequential risk Correlation across risk Correlated risks are difficult to model in decision trees Risk type and Probabilistic Approaches 79-79 Discrete/ Continuous Correlated/ Independent Sequential/ Concurrent Risk approach Discrete Correlated Sequential decision trees Discrete Independent Concurrent scenario analysis Continuous Either Either simulations