Transcript

00:01 |

Initial credits
00:06 |

Introduction
01:17 |

Standard nonparametric tests
- Abnormal distribution in tests
- More flexibility in tests
04:22 |

Nonparametric tests
05:40 |

Paired and unpaired comparison
06:45 |

Wilcoxon-Mann-Whitney test
- History of the test
- Key assumptions required
- Conditions in which assumptions are violated
- Independent variables
- Test statistic
16:40 |

Example of a Wilcoxon-Mann-Whitney test (Uri Gneezy and John A. List, 2006)
- Interpretation of the independent observations
- Simplicity of statistics
- Comparison between treatments
- Scenario of one treatment consistently higher than the other
- Scenario of treatments consistently the same
- Distribution of Wilcoxon-Mann-Whitney test statistics
- Asymptotic properties of Wilcoxon-Mann-Whitney test
30:44 |

Nature of asymptotic distribution
32:01 |

Nonparametric test and the data needed
34:19 |

Median test
- Degrees of freedom
- Pros and cons of Median test
39:20 |

Wilcoxon signed-rank test
- Example of a Wilcoxon signed-rank test
- Interpretation of the results of the example test
- Asymptotic properties of Wilcoxon signed-rank test
45:58 |

Jonckheere-Terpstra test
One-sided test statistic
48:53 |

Page test
49:22 |

Adaptive procedures for nonparametric tests
50:20 |

Most powerful rank test
53:15 |

Cons of the most powerful rank test
54:56 |

Correct use of the data
57:34 |

Gastwirth's modified rank test
58:38 |

Differences between uniform distributions
59:20 |

Symmetric and skewed distribution
01:01:36 |

Adaptive distribution-free procedure
01:02:03 |

Hogg, Fisher and Randles' (HFR) skewness and tailweight
- Model selection scheme
- Choice of modified tests
01:05:58 |

Optimized algorithm 1: Model selection scheme
Optimization of the cup points
01:09:02 |

Optimized algorithm 2: Choice of modified tests
01:09:31 |

Monte-Carlo analysis of power and size
01:10:49 |

Comparison between HH, MWW and t-test
- Results of the t-test
- Importance of statistical significance
01:15:04 |

Light-tailed and right-skewed sample test
01:16:38 |

Medium and heavy-tailed sample test
01:17:30 |

Empirical size and power
01:18:30 |

Adaptive procedures
01:18:52 |

When to use adaptive procedures
01:19:54 |

Recommendations for binary data
01:22:10 |

Final credits

Nonparametric Statistics for Experiments

New Media | 23 de octubre de 2012 | Vistas: 354

Experimental Economics

Daniel Houser, a researcher on experimental statistics and methods, provides along his lecture, details on nonparametric tests and their significance and application in social and economic analysis.

He shares an explanation on nonparametric tests, also known as "distribution-free tests," which are statistics that do not rely on the data belonging to any particular distribution, and also discusses the special features of these techniques. The conference is divided in two parts, the first being the Standard Nonparametric Tests and the second, Adaptive Procedures for Nonparametric tests.

Houser gives an overview of the most important statistical hypothesis tests and a comparison between one another in sample tests. For each, an explanation on how to interpret the results is presented, as well as the best scenario for the test to be applied, being among them the Wilcoxon-Mann-Whitney test, Median test, Wilcoxon signed-rank test, Jonckheere-Terpstra test, Page test, Gastwirth's modified test. He concludes by giving recommendations for analysis of binary data tests.