School Projects

Take a look at what I learned in my statistics courses. I understand a variety of statistical methods to make predictions or infer causation from natural experiments and imperfect data. I also know how to read regression results, specify models, and employ statistical methods to deal with endogeneity and heteroskedasticity.

DIDgraph.png

Differences-in-Differences

We used data collected by David Card and Alan Krueger on fast-food restaurants in New Jersey and eastern Pennsylvania during two interview waves in March and Nov/Dec of 1992. On April 1, 1992, New Jersey raised its minimum wage from $4.25 to $5.05. The minimum in Pennsylvania remained at the federal level of $4.25. We used this data to analyze the impact of the minimum wage increase in New Jersey on employment in the fast-food industry.

Unbalanced Panel

Unbalanced panel data on 195 countries from 1960 - 2000. From "Income & Democracy" By Prof. Daron Acemoglu, Published in American Economic Review, 2008, 98:3: 808-842. 

Hausman test, F-tests, specification to determine if 

icarifpri-instrumental-variable-regressi

IV Regression

We used a dataset based on a study by Acemoglu, Johnson, and Robinson (AJR). The dataset contains information about 62 non-European countries. We tested the hypothesis assessed by AJR that the protection of property rights is conducive to growth, and is correlated with the level of contemporary per capita income. 

FIXED-EFFECT+MODEL+RANDOM-EFFECTS+MODEL.

Fixed / Random Effects

The effect of guns on crime. Balanced panel data on 50 US states and DC from 1977 - 1999. From "Shooting Down the 'More Guns Less Crime' Hypothesis" by Prof. John Donohue, Published in Stanford Law Review, 2003, Vol. 55, 1193 - 1312. 

We tested whether the year dummies jointly equal zero to decide if we should use year fixed effects in the model. Then, we used the Wu-Hausman test to decide between a fixed effects or a random effects model.

psmatching.png

Propensity Score Matching

We used a dataset based on a study on maternal smoking and birthweight to work through an example of propensity score matching.  We compared the results from OLS to teffects.  We calculated both the average treatment effect and the average treatment effect on the treated and saved the results using estimates store and then tabulated all three results for comparison.

Plotting Beta

Used data on the dividend-adjusted price of 10 stocks and the S&P 500 index and the risk-free rate from Yahoo  Finance.

We collapsed and merged data sets and measured the alpha and beta of major stocks.