CFA Level II Without Review and Secret Sauce Flashcards
All areas
What is R2?
SST-SSE/SST or RSS/SST
Adjusted R2
1- [((n-1)/(n-k-1)) x (1-R2)]
What is the difference between AIC and BIC
AID is for a forecast and BIC is goodness of fit a lower number is better
What is the F statsitic
(SSER - SSEu)/q / (SSEu)/(n - k - 1) with q and (n-k-1) degrees of freedom q = number of excluded variables in the restricted model and k = independent variables in the full model
What is the BP Chi-square test statistic
n x R2resid with k degrees of freedom where n is the number of observations, and R2 from a second egression on the independent variables and k is the number of independent variables.
What is the BG test
The BG test regresses the regression residuals against the original set of independent variables plus one more additional variables representing lagged residuals.
What is the Variance Inflation Factor?
It is used to quantify multicollinearity where VIF - 1 / (1-R2J) high VIF is what we are looking for
What is Cook’s Distance
It is a composite metric to evaluate if an observation is influential. It is D = ei2/(k+1) x MSE [hii/(1 - hii)^2]
What is the logistic regression
ln(p/1-p) = b0 + b1X1 + b2X2
What are odds?
Odds = e^y and P = odds/(1 + odds) = 1/(1 + e^-y)
What is an autoregressive model?
An AR is when the dependent variable is regressed against one or more lagged values of itself
What is the Durbin Watson or DW statistic?
It is used to detect autocorrelation and if the timeseries does not have autocorrelation it should be 2.0
How to forecast with AR models?
First start with the first orderAR model and then you find the autocorrelation of the residuals. After that you see if they are different from zero.
What is the formula for mean reversion for a linear regression?
xt = b0 / (1-b1)
What is a random walk with a drift?
It is when in addition to a random error the time series is expected to increase or decrease by a constant amount each period where xt = b0 + b1xt-1 + et where b0 = the constant drift and b1= 1
How do we test for random walk and when do we have unit root?
If the value of the lag coefficient is equal to one the time series is said to have unit root and follow a random walk
What is the Dickey Fuller aka Engle Granger test?
Used to test for unit root and it tests is an AR times series is equal to 1.
What is Autoregressive conditional heteroskedasticity (ARCH)
It exists if the variance of the residuals in one period is dependent on the variance of the residuals in a previous period and is used to test for AR conditional heteroskedasticity where et^2 = a0 + a1e^2t-1 + meant
What is cointegration?
It is when two time series are economically linked (related to the same macro variable) or follow the same trend and that relationship is not expected to change.
What is the difference between supervised and unsupervised learning?
Supervised uses labeled training data while unsupervised are not given labeled training data.
What is bias and variance error and how do they change with model complexity?
Bias error is the in-sample error resulting from a model with poor fit while variance is the out of sample error resulting from overfitting models that do not generalize well. Variance increases with model complexity and bias decreases.
What are eigen vectors?
They are uncorrelated factors which are a combinations of the original features and it the proportion of total variance in the data set explained. This is key in Principal component analysis
What is the difference between trimming and winorizing
Trimming is where the highest and lowest x% of observations are removed and winorizing is where extreme values are placed by the maximum value
What is normalization and how do you calculate it?
Normalization scales the variable between 0 and 1. Normalized X = (Xi - Xmin)/(Xmax - Xmin)