Econometrics with Eviews: Time Series, and Panel Data

Learn to convert time series data into panel data by assembling country data (Pakistan, India, Bangladesh, Sri Lanka) for inflation, GDP, and FDI, using cross section IDs and start-end dates.

Open EViews, create a new workfile, set the data form and dates, then import data from Excel by copying or by importing from a file.

Learn how to compute descriptive statistics in EViews for panel data and time series, including mean, median, max, min, standard deviation, skewness, kurtosis, Jarque-Bera, and sums across 64 observations.

Use the Ramsey reset test to diagnose whether the IV and DV relationship is linear, quadratic, or cubic. Interpret p-value at 0.05 to judge if the model is correctly specified.

Learn to create lags in Eviews, including one lag and two lags, using generate and the command window, with GDP lag one and GDP lag two.

Learn to select the optimal lag in an EViews var by comparing criteria (LR, FPE, AIC, SC, HQ) and choosing the lag supported by the majority of tests.

Identify outliers using leverage and influence diagnostics in EViews after running a simple OLS, then transform data (growth rates or logs) to reduce bias for regression.

Learn how MIDAS enables mixed frequency data analysis by modeling low frequency dependent variables with high frequency regressors, preserving information for nowcasting and forecasting, with practical steps in EViews.

Explore the assumptions of the classical linear regression model for reliable OLS results, including linearity in parameters, variable x, x non-stochastic, homoscedasticity, zero conditional mean, exogeneity, and correct specification.

Apply the first OLS assumption: linearity in parameters enables estimation; nonlinear in parameters blocks estimation unless you transform to a linear form.

Explains homoscedasticity as the constant variance of the error term across all levels of the independent variable in the classical linear regression model, and contrasts it with heteroscedasticity.

Understand zero conditional mean of residuals in the classical linear regression model, with residuals random and uncorrelated with predictors, and identify violations like omitted variables, measurement error, and model specification.

Understand why in ordinary least squares regression, the number of observations must exceed the number of parameters to ensure unique, reliable estimates and avoid underdetermined models.

Ensure regression models are correctly specified with the right functional form and relevant variables. Avoid irrelevant variables and measurement error to prevent bias, inefficiency, and incorrect conclusions.

Compare simple and multiple linear regression, where simple uses one independent variable and multiple uses two or more to explain income, with an intercept, coefficients, and an error term.

Derive the ordinary least squares estimator for a two-variable multiple regression with x1 and x2, using mean centering and normal equations to obtain beta1, beta2, and beta0.

Compute the multiple regression estimators beta hat one, beta hat two, and beta hat naught from a numerical example, using means, sums, and cross products with x1, x2, and y.

Explore identifying multiple structural breaks using the Bai-Perron test in EViews, by importing data, selecting GDP and inflation, and running stability diagnostics for break points.

Learn practical methods to detect heteroscedasticity in regression using residual diagnostics and tests like Breusch-Pagan and White, interpret p-values, and assess data quality.

Explore how heteroscedasticity makes OLS estimates unbiased and consistent but inefficient due to larger variances, inflating standard errors and weakening t and F tests.

Explore remedial measures for heteroscedasticity in regression analysis, including weighted least squares, generalized least squares, variable transformations, and model specification to achieve homoskedasticity.