The effect of interaction and rounding error in two-way ANOVA: example of impact on testing for normality

被引:5
作者
Ryden, Jesper [1 ]
Alm, Sven Erick [1 ]
机构
[1] Uppsala Univ, Dept Math, S-75238 Uppsala, Sweden
关键词
analysis of variance; rounding error; Shapiro-Wilk test; factorial experiment; interaction; VALUES;
D O I
10.1080/02664760903143925
中图分类号
O21 [概率论与数理统计]; C8 [统计学];
学科分类号
020208 ; 070103 ; 0714 ;
摘要
A key issue in various applications of analysis of variance (ANOVA) is testing for the interaction and the interpretation of resulting ANOVA tables. In this note it is demonstrated that for a two-way ANOVA, whether interactions are incorporated or not may have a dramatic influence when considering the usual statistical tests for normality of residuals. The effect of numerical rounding is also discussed.
引用
收藏
页码:1695 / 1701
页数:7
相关论文
共 13 条
  • [1] [Anonymous], APPLIED STATISTICS
  • [2] GJEDDEBAEK NF, 2000, INT ENCY STAT, P1056
  • [3] ANALYSIS OF A 2-WAY MODEL WITH INTERACTION AND NO REPLICATION
    JOHNSON, DE
    GRAYBILL, FA
    [J]. JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1972, 67 (340) : 862 - 868
  • [4] Montgomery D., 2010, Design and Analysis of Experiments
  • [5] The impact of data accuracy on the POT estimates of long return period design values
    Naess, A
    Clausen, PH
    [J]. JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING-TRANSACTIONS OF THE ASME, 2002, 124 (01): : 53 - 58
  • [6] PEARSON ES, 1977, BIOMETRIKA, V64, P231
  • [7] PREECE DA, 1982, J ROY STAT SOC D-STA, V31, P169
  • [8] RYDEN J, 2006, P 16 INT OFFSH POL E, V3, P81
  • [9] SHEPPARD WF, 1898, P LOND MATH SOC, V29, P353, DOI 10.1112/plms/s1-29.1.353
  • [10] Speed TP, 2002, ANN STAT, V30, P1649