CFA Level II QM Quizlet

Name: Score: 45 Multiple choice questions Definition 1 of 45 Failed to reject a false null hypothesis. False negative: negative prediction, but positive actual. Variance error Type I error Type II error Standard error Definition 2 of 45 (TP + TN) / (all P + all N). Minimizes overall Type I and II errors Regulation Accuracy Occurrence Accessibility Definition 3 of 45 regressing squared resid on previous squared resid. If coef significantly different from 0, then ARCH exists, in which case use generalized least squares to modify COLUMN ARCH DELTA RADIAL Definition 4 of 45 Appropriate to apply when variable growth at a concentrate over time Logarithmic transformation Standardization Cointegration Square root transformation Definition ε 5 of 45 xt = b0 + b1xt-1 + ... + t BIC AR model MSR ARCH Definition Is a problem. Variance of error correlated with independent variables. F-stat unreliable. T-stat inflated because standard errors are understated. Coefficient still consistent and unbiased. Unconditional heteroskedasticity Multicollinearity Autoregressive heteroskedasticity Conditional heteroskedasticity 6 of 45 Definition 7 of 45 Heteroskedasticity in residuals cointegration accuracy white corrected standard errors inappropriate transformation Definition x̄ μ √ x̄ μ 8 of 45 t = ( - ) / (s / n) = ( - ) / SE t-test t-value correlation coefficient type i error Definition 9 of 45 [(SSE_R - SSE_U)/q] / [SSE_U / (n-k-1)] q = number of restrictions (count equal signs. H=0: B0 + B1 = 1 yield q of 1 while H0: B0 = B1 = 0 yields a of 2) reject H0 if F(test-stat) > F(critical value) Serial correlation Dickey-Fuller Test F-test unrestricted F-test restricted Definition 10 of 45 To correct for serial correlation in residuals Newey-West corrected standard errors Data improperly pulled Type I error White corrected standard errors Definition 11 of 45 No major problems Unconditional heteroskedasticity Data improperly pulled Cointegration White corrected standard errors Definition 12 of 45 Heteroskedasticity / serial correlation in residuals White corrected standard errors F-test restricted F-test unrestricted Data improperly pulled Definition If both timeseries have unit roots, they must exhibit coin integration for the results of the regression to be valid Cointegration Covariance Stationary Autocorrelation Omission Of Important, Independent Variables 13 of 45 Definition 14 of 45 Test for autocorrelation in regression residuals. Residuals regress against X variables and lagged residuals. H0: no serial correlation. Durbin-watson Test Breusch-Godfrey Test Dickey-Fuller Test Jarque-bera Test Definition 15 of 45 TP / (TP + FN) Higher recall = lower Type II error Type II Error Accuracy Recall Recognition Definition 16 of 45 Positive autocorrelation. T-stat inflated (Type I error) because standard errors are understated. Coefficients still consistent and unbiased. Joint regression Serial correlation Inappropriate transformation Accuracy Definition 17 of 45 condition for using AR models 1. Expected value is constant and finite= mean reverting. 2. Variance is constant and finite. 3. Covariance is constant and finite. Variance inflation factor (VIF) Covariance stationary Dickey-Fuller Test Serial correlation Definition 18 of 45 RSS/k MSE VIF MSR MSC Definition -Linearity: Linear relationship between Y and X -Expected value of error term is zero -Normality: Error term normally distributed -Homoskedasticity: Variance of error term is constant (no heteroskedasticty) -Errors are not serially correlated (no autocorrelation) -No exact linear relationship among X (no multicollinearity) Assumptions of multiple regression Conditional heteroskedasticity Unconditional heteroskedasticity Variance inflation factor (VIF) 19 of 45 Definition 20 of 45 MRL = b0 / (1-b1) AR(1) model mean reversion Unsupervised Learning High leverage point White corrected standard errors Definition 21 of 45 Extreme Y variable value Cluster Slope Standard Deviation Outlier Definition 22 of 45 To correct for conditional heteroskedasticity White corrected standard errors Data improperly pulled Type I error Newey-West corrected standard errors Definition Neural networks NLP reinforced learning Machine Learning Previewing Programmed Learning Deep Learning 23 of 45 Definition 24 of 45 TP / (TP + FP). Higher Precision = Lower Type I error Type I Error Accuracy Repeatability Precision Definition 25 of 45 RSS/TSS Mean Squared Error R^2 R^0 Standard Error Definition Extreme X variable value High leverage point High residual point Type I error Accuracy 26 of 45 Definition 27 of 45 T-stat understated because standard errors are inflated, unreliable coefficients (but still consistent) higher chance of Type II errors Linearity Multicollinearity Micronumerosity Homoskedasticty Definition 28 of 45 Detects multicollinearity. VIF_j = 1 / (1-R_j). VIF > 5 warrants further investigation. VIF > 10 indicates serious multicollinearity. Both a and b Variance inflation factor (VIF) Correlation coefficient (r) Multiple r (correlation coefficient) Definition 29 of 45 False rejection of a true null hypothesis. false positive: positive prediction but negative actual Standard error Type I error Type II error Sampling error Definition 30 of 45 [(SST_U - SSE_U)/k] / [SSE_U / ( n-k-1)] = MSR / MSE T-test unrestricted Chi-square test unrestricted F-test unrestricted Z-test unrestricted Definition 31 of 45 best for goodness of fit, lower is better nln(SSE/n) + ln(n)*(k+1) MSR RCM AICC BIC Definition 2PR / (P + R) F1 score Accuracy score Recall score Precision score 32 of 45 Definition 33 of 45 best for prediction, lower is better. nln(SSE/n) + 2(k+1) AIP MSR AIH AIC Definition 34 of 45 For unit root. ie coefficient = 1. Having a unit root means random walk and non-stationary Do first differencing: x_t - x_[t-1] = b0 + (b1-1)x_[t-1]. If there is a unit root then g=(b1-1)=0 H0: g=0 Dickey-Fuller Test Covariance stationary F-test restricted Serial correlation Definition Heteroscedasticity in residuals, multicollinearity Data improperly pulled Omission of important, independent variables High leverage point Inappropriate variable scaling 35 of 45 Definition 36 of 45 Test for autocorrelation in regression residuals. Only applies to first order serial correlation Durbin-Watson statistic Breusch-Godfrey Test Breusch-pagan test Dickey-Fuller Test Definition 37 of 45 Biased and inconsistent regression parameters, heteroskedasticity / serial correlation in residuals Inclusion of irrelevant variables Multicollinearity among independent variables Omission of important, independent variables Transformation of dependent variable Definition Cannot use Durbin-Watson. Testing for autocorrelation in MA model Testing for heteroskedasticity in time Series model Testing for surreal correlation in AR model Testing for multicollinearity in regression model 38 of 45 Definition 39 of 45 1 / [1 + e^(-y)] Accuracy AR model Logistic model probability Type I error Definition 40 of 45 SSE/(n-k-1) AIC MSR SDS MSE Definition 41 of 45 Chi-Squared test for heteroskedasticity where the squared OLS residuals are regressed on the X variables and test for significance of resulting R^2: BP = (R_resid)^2 x n H0: No heteroskedasticity ARCH MSE Durbin-Watson Statistic Breusch-Pagan Definition 42 of 45 PCA K-means clustering Data Mining Supervised Learning Machine Learning Unsupervised Learning Definition Ln[p/(1-p)] = b0 + b1X1 + ... + ε 43 of 45 Exponential model Geometric and exponential growth model Logistic model AR model Definition 44 of 45 1-[(n-1)/(n-k-1)*(1-R^2)] Precision AR model Outlier R^2_adj Definition Lasso SVM KNN cart Supervisor learning Type I error Deep Learning AR model 45 of 45