Results

**1 - 7**of**7**### Repository CRAN

, 2010

"... Depends maps Description Provides mailmerge methods for reading spreadsheets of addresses and other relevant information to create standardized but customizable letters. Provides a method for mapping US ZIP codes, including those of letter recipients. Provides a method for parsing and processing htm ..."

Abstract
- Add to MetaCart

Depends maps Description Provides mailmerge methods for reading spreadsheets of addresses and other relevant information to create standardized but customizable letters. Provides a method for mapping US ZIP codes, including those of letter recipients. Provides a method for parsing and processing html code from online job postings of the American Political Science Association.

### Supplementary Materials for “Multivariate Continuous Blocking to Improve Political Science Experiments”⇤

, 2013

"... Consider a small political field experiment in which you will assign a campaign message in four precincts, two predominantly Democratic and two predominantly Republican. To prevent your preconceptions from influencing the experiment, you decide to randomize the two available campaign messages (say, ..."

Abstract
- Add to MetaCart

Consider a small political field experiment in which you will assign a campaign message in four precincts, two predominantly Democratic and two predominantly Republican. To prevent your preconceptions from influencing the experiment, you decide to randomize the two available campaign messages (say, A and B) to the precincts. You know that precinct partisanship is likely to strongly a↵ect voters ’ responses to the campaign messages. You randomly select two precincts to assign message A to, and both Democratic precincts are chosen. What do you do? If you proceed, you will be unable to untangle the e↵ects of partisanship from that of the campaign messages. Better design could have prevented this problem. By simply sorting the precincts by partisanship then randomizing one Democratic and one Republican precinct to each message, the experiment could generate more useful information. This pre-randomization sorting of units, blocking, can easily incorporate many variables, whether discrete or continuous, and confers several statistical and political ad-

### License GPL (> = 2) | file LICENSE

, 2014

"... Description Blocks units into experimental blocks, with one unit per treatment condition, by creat-ing a measure of multivariate distance between all possible pairs of units. Maximum, mini-mum, or an allowable range of differences between units on one variable can be set. Ran-domly assign units to t ..."

Abstract
- Add to MetaCart

Description Blocks units into experimental blocks, with one unit per treatment condition, by creat-ing a measure of multivariate distance between all possible pairs of units. Maximum, mini-mum, or an allowable range of differences between units on one variable can be set. Ran-domly assign units to treatment conditions. Diagnose potential interference between units as-signed to different treatment conditions. Write outputs to.tex and.csv files.

### Political Science 590 Political Experiments: Design & Analysis, Part I

, 2012

"... General Information Overview Randomized interventions allow political scientists to claim that comparisons are causal: randomization allows us the ability to characterize counterfactual comparisons- to say how the treated group would have responded had treatment been withheld. Randomization also all ..."

Abstract
- Add to MetaCart

General Information Overview Randomized interventions allow political scientists to claim that comparisons are causal: randomization allows us the ability to characterize counterfactual comparisons- to say how the treated group would have responded had treatment been withheld. Randomization also allows us to test hypotheses about causal comparisons without requiring large samples or probability models of outcomes. Randomized experiments thus promise to simplify our lives and to enable clear answers to important substantive and theoretical questions without requiring much in the way of extraneous justification. Yet, there is an art in designing and analyzing a randomized experiment. Only the simplest of experiments can be analyzed simply. And, in political science, we may not be able to directly require subjects to be exposed to a dose of our treatment and may need to work indirectly, using the randomization as an instrument to manipulate a dose, perhaps at a distance, perhaps weakly. In this course, we will practice designing experiments and making statistical inferences from the resulting data. Building on the most basic foundations (what is an experiment? why randomize?) we will engage with questions about clustered assignment, blocking, multi-valued treatments, and instrumental variables. Our hope is to prepare students to face their own design and analysis decisions of lab and field experiments with creativity and with a strong foundation in the design and statistical literatures on these topics. Over the two terms of this course and the winter break, we anticipate that you will design, field, and analyze data from, a randomized experiment. This course is brand new. Expect changes to the syllabus and format of the course as we go along. We welcome constructive comments and suggestions.

### Improving Experiments by Optimal Blocking: Minimizing the Maximum Within-block Distance

"... We develop a new method for blocking in randomized experiments that works for an arbitrary number of treatments. We analyze the following problem: given a threshold for the minimum number of units to be contained in a block, and given a distance measure between any two units in the finite population ..."

Abstract
- Add to MetaCart

We develop a new method for blocking in randomized experiments that works for an arbitrary number of treatments. We analyze the following problem: given a threshold for the minimum number of units to be contained in a block, and given a distance measure between any two units in the finite population, block the units so that the maximum distance between any two units within a block is minimized. This blocking criterion can minimize covariate imbalance, which is a common goal in experimental design. Finding an optimal blocking is an NP-hard problem. However, using ideas from graph theory, we provide the first polynomial time approximately optimal blocking algorithm for when there are more than two treatment categories. In the case of just two such categories, our approach is more efficient than existing methods. We derive the variances of estimators for sample average treatment effects under the Neyman-Rubin potential outcomes model for arbitrary blocking assignments and an arbitrary number of treatments.

### Estimation of treatment effects

, 2013

"... Blocking and graph theory Approximately optimal blocking algorithm ..."

(Show Context)
### Improving Experiments by Optimal Blocking: Minimizing the Maximum Within-block Distance

"... We develop a new method for blocking in randomized experiments that works for an arbitrary number of treatments. We analyze the following problem: given a threshold for the minimum number of units to be contained in a block, and given a distance measure between any two units in the finite population ..."

Abstract
- Add to MetaCart

(Show Context)
We develop a new method for blocking in randomized experiments that works for an arbitrary number of treatments. We analyze the following problem: given a threshold for the minimum number of units to be contained in a block, and given a distance measure between any two units in the finite population, block the units so that the maximum distance between any two units within a block is minimized. This blocking criterion can minimize covariate imbalance, which is a common goal in experimental design. Finding an optimal blocking is an NP-hard problem. However, using ideas from graph theory, we provide the first polynomial time approximately optimal blocking algorithm for this problem. We derive the variances of estimators for sample average treatment effects under the Neyman-Rubin potential outcomes model for arbitrary blocking assignments and an arbitrary number of treatments.