讲解 Quantitative Analysis II讲解 Java语言

The following topics are covered in Quantitative Analysis II and are necessary background for Quant III courses.Since the Proficiency Exam is an open book exam, students may bring any econometrics statistics textbook with which they are comfortable.

Multiple Regression I: Dummy Variables and Interactions (this is just carry-over from Quant)

· Categorical Independent Variables

o Understand the concept of the base (omitted) category

o Be able to interpret b-coefficients for a set of (J - 1) dummy variables

· Interaction terms

o Understand and be able to interpret an interaction term that is the product of one or more dummies and a single continuous regressor

o Understand and be able to interpret interaction terms that is a product of two dummy regressors

Specification I: Choosing the Right Variables

· Multiple Regression

o Understand and be able to interpret standardized betas

o Understand and be able to calculate and interpret the Adjusted R-square

· Omitting a Relevant Independent Variable

o Understand the consequences of omitting a relevant independent variable

o Be able to anticipate the direction of omitted variable bias in an application

· Including an Irrelevant Independent Variable

o Understand the consequences of including an irrelevant independent variable

Specification II: Correct Functional Form

· Linear Functional Form.

o Understand and be able to diagnose violations of linearity from scatterplot output

· Linear-Log Models

o Understand when you might want to use this model

o Be able to interpret the slope coefficients (using the approximation)

o Be able to interpret the b-coefficient for a logged X predictor by calculating consecutive predicted values and assessing the predicted change in Y that results from a change in X

· Log-Linear Models

o Understand when you might want to use this model

o Be able to interpret the slope coefficients (using the approximation)

· Log-Log Models

o Understand when you might want to use this model

o Be able to interpret the slope coefficient as an elasticity

o Be able to calculate the expected value of the slope of curve any specific value of X_i′

· Polynomial Model (Quadratic)

o Understand when you might want to use this model (based on theory and scatterplot)

o Understand and be able to solve for and interpret the value of X_i′ at the turning point

o Be able to calculate and interpret the slope of the curve at any value of X_i′

o Be able to calculate and interpret “marginal effects” of X on Y (i.e., the predicted change in Y for a one-unit change in X--at different levels of X—using predicted values)

Multicollinearity

· Meaning of Multicollinearity

o Understand multicollinearity as a sample phenomenon

o Understand the practical consequences of multicollinearity

· Detecting Multicollinearity

o Informally (compare individual t-values to measures of overall model fit, simple bivariate correlation coefficients)

o Understand and be able to calculate and interpret Variance Inflation Factor

· Correction Strategy for Multicollinearity

o Understand and be able to evaluate and interpret possible correction strategies when given relevant information (i.e., word problem/description, STATA output)

Heteroskedasticity

· Meaning of Heteroskedasticity

o Understand the meaning of (pure) heteroskedasticity

o Understand the practical consequences of heteroskedasticity

· Detecting Heteroskedasticity

o Informally (plot of residuals against X_i’s, plot of residuals against predicted values)

o Understand the motivations of the Breusch-Pagan and White tests

§ Understand the form. of heteroskedasticity being tested in each test

o Be able to execute these tests and interpret their results

§ Understand and be able to apply the appropriate decision rule

· Robust Standard Errors

o Understand the motivation and calculation of robust standard errors

o Understand and be able to interpret regression results that use robust standard errors

Serial Correlation

· Understand the problem of first-order serial correlation

· Understand the practical consequences of first-order serial correlation

· Detection of first-order serial correlation

o Be able to perform. the Durbin-Watson d test

· Robust Standard Errors

o Understand the motivation behind robust standard errors

o Understand and be able to interpret regression results that use Newey-West robust standard errors

Linear Probability Model and Logit

· Linear Probability Model

o Understand the motivation and meaning of a binary dependent variable model

o Understand the inherent problems with the LPM

o Be able to interpret slope coefficients in a LPM

· Binary Logistic Regression

o Be able to calculate and interpret logit coefficients as odds ratios

o Be able to calculate and interpret predicted probabilities (using the formula) and predicted probability changes (using relevant STATA output from margins/marginsplot)

o Understand and be able to calculate and interpret the percentage of cases correctly predicted (using lstat)

Difference in Differences

· Understand the basic DID setup as a comparison of four sample means (i.e., Y’s)

· Understand a DID setup with two independently pooled cross-sections

· Understand a DID setup with two period of panel data on the same cross-sectional units

· Be able to calculate and interpret the DID estimate in either of the two above setups

· Understand the key identifying DID assumption (“common trends”), and how one might demonstrate emirically the plausibility of this assumption

Instrumental Variables

· Understand the idea of the four compliance types using a potential outcomes framework

· Understand the econometric formulation of the IV estimand (two composite assumptions)

· Understand and be able to compute the IV estimand as the ratio of two causal effects (i.e., the Wald estimate)

o Understand the numerator (ITT)

o Understand the denominator (causal effect of Z on D)

· Be able to interpret 2SLS regression estimates (with and without covariates)

· Understand how 2SLS can be extended to incorporate multiple instruments

· Understand how 2SLS can be extended to incorporate a continuous treatment (D)

Panel Data Methods

· Understand how pooled OLS can be used when analyzing panel data (and what its limits are)

· Understand how panel data can help with the problem of omitted variable bias

· Understand how we control for unit fixed effects and time fixed effects

· Understand how to assess between and within variation on the treatment variable

· Understand how first-differencing eliminates the unit fixed effects prior to estimation

o Understand how to interpret the coefficient estimates in first-differences models

· Understand how to apply and interpret the deviations from the means estimator

o Understand how to interpret the coefficient estimates

o Understand xtreg, fe output (including post-estimation)

· Understand how to apply and interpret the LSDV estimator

o Understand how to interpret the coefficient estimates (included the estimated fixed effects)

· Understand the key identifying assumptions behind fixed effects regression models