W2D2S2: Social Networks & College Performance/ Power Curve Simulation Study

Social Networks and College Performance: Evidence from Dining Data, John Krieg (Professor of Economics, Western Washington University)

We investigate the effect of friends on academic performance in college using unique data on dining card swipes at a medium-sized public university. Using several different time- and frequency-bandwidth measures, we define friendships by academic quarter as repeated meetings among students in the same dining hall. To identify the impact of having a friend in class, we employ models using student- and class-level fixed effects as well as a number of controls to rule out alternative explanations. Our results suggest that having a friend in class has a large and positive effect on grades, and this effect is consistent regardless of a friend’s background characteristics.

Power Curve Simulation Study of Nonparametric Multiple Contrast Testing Procedures for One-Way Repeated Measures Experimental Design. Patrick Caroll (Oregon State University)

Repeated measures experimental designs frequently arise in many fields of study, from medicine to psychology to policy making, where multiple observations are taken from the same set of subjects. The use of rank-based nonparametric methods is commonly recommended for analyzing effects in these designs as the observed data typically have non-normal distributions with possibly a number of outliers.

In this study, the sizes and powers of the recent nonparametric multiple contrast testing procedures for the one-way repeated measures experimental design are compared empirically. The methods included in this study are the multivariate normaland t-based methods implemented in the mctp.rm() function of the nparcomp R package. In addition, the df-procedure presented in Hasler (2013) which suggests an alternative way of computing the degrees of freedom for the multivariate t-based method is examined.