Exact and Pseudo P-values in the Wilcoxon Unpaired Test with Ties

スポンサーリンク

概要

• 論文の詳細を見る

• This paper investigates the discrepancy between the exact and the pseudo p-values of the Wilcoxon unpaired test statistics for several tie patterns. The findings are summarized as follows. First, the discrepancy between the exact and the pseudo p-values tends to be less than 5% when the sample size is greater than 15 for each sample set and the size of the ties takes 15% at most in the sample. Secondly, the exact and the pseudo p-values are very similar at the significance levels from 0.01 to 0.1. Since these levels are frequently used in practice, the statistical testing is less likely to be misleading. Thirdly, existence of mid ranks or a large size of the ties beyond 20% in the sample deteriorates approximations by the pseudo p-values. Finally, the exact p-values are easily obtained with a simple algorithm on a worksheet application when each sample size of two data groups is less than or equal to 16.

• 成城大学の論文

著者

• 大津 武

  成城大学経済学部

関連論文

• Separating Trends and Cycles
• Seasonal Cycle and Filtering
• Exact and Pseudo P-values in the Wilcoxon Unpaired Test with Ties
• Time-Invariant Linear Filters and Real GDP : A Case of Japan
• Stability of Extracted Cycles : A Case of Adjustment Factors in Quarterly Data
• Toward Harmless Detrending
• Peaks and Troughs of Cycles and Filtering Methods
• Time-Invariant Linear Filters and Real GDP : A Case of Japan

スポンサーリンク