The accuracy of four approximations to noncentral F

被引:3
作者
Bradley, DR [1 ]
Russell, RL
Reeve, CP
机构
[1] Bates Coll, Dept Psychol, Lewiston, ME 04240 USA
[2] Westinghouse Savannah River Co, Aiken, SC 29808 USA
来源
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS | 1998年 / 30卷 / 03期
关键词
D O I
10.3758/BF03200682
中图分类号
B841 [心理学研究方法];
学科分类号
040201 ;
摘要
The accuracy of the two-moment, three-moment, square root, and cube root approximations to the noncentral F distribution was assessed using 7,920 entries from Tiku's (1967) power tables. Tiku's tables list exact values of beta for alpha =.005-.05, nu(1) = 1-12, nu(2)= 2-120, and phi = 0.5-3.0. Analysis of the errors showed generally satisfactory performance for all four approximations. The three-moment approximation was most accurate, registering a maximum error of only .009. The other three approximations had maximum errors of +/-.02, except for the square root approximation at nu(2) = 2, where maximum errors of .05 occur. Approximation error increased with decreases in nu(1) and, less consistently with increases in nu(2). Error was nonmonotonically related to phi. A second investigation explored the accuracy of the approximations at values of alpha ranging from .10 to .90. All four approximations degraded substantially in this situation, with maximum errors ranging from -.09 to .05. If the analysis is restricted to cases where nu(1) > 1 and nu(2) > 2, maximum errors drop to roughly +/-.03. We conclude that the approximations perform reasonably well for small a and moderately well for larger values, if certain restrictions are imposed. From a computational standpoint, however, there is little advantage to using approximate as opposed to exact methods unless exact values of F-alpha are known in advance.
引用
收藏
页码:478 / 500
页数:23
相关论文
共 55 条
[1]  
ABRAMOWITZ M, 1964, APPL MATH NATL BUREA, V55
[2]  
BARTON DE, 1960, BIOMETRIKA, V47, P417
[3]  
Borenstein M., 1988, Statistical Power Analysis: A Computer Program
[4]   COMPUTER-SIMULATION WITH DATASIM [J].
BRADLEY, DR .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1989, 21 (02) :99-112
[5]   STATISTICAL SIMULATION ON MICROCOMPUTERS [J].
BRADLEY, DR ;
SENKO, MW ;
STEWART, FA .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1990, 22 (02) :236-246
[6]   Statistical power in complex experimental designs [J].
Bradley, DR ;
Russell, RL ;
Reeve, CP .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1996, 28 (02) :319-326
[7]   ANATOMY OF A DATASIM SIMULATION - THE DOOB AND GROSS HORN-HONKING STUDY [J].
BRADLEY, DR .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1991, 23 (02) :190-207
[8]   MULTIVARIATE SIMULATION WITH DATASIM - THE MIHAL AND BARRETT STUDY [J].
BRADLEY, DR .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1993, 25 (02) :148-163
[9]   GENERATING MULTIVARIATE DATA FROM NONNORMAL DISTRIBUTIONS - MIHAL AND BARRETT REVISITED [J].
BRADLEY, DR ;
FLEISHER, CL .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1994, 26 (02) :156-166
[10]   A SIMULATION LABORATORY FOR STATISTICS [J].
BRADLEY, DR ;
HEMSTREET, RL ;
ZIEGENHAGEN, ST .
BEHAVIOR RESEARCH METHODS INSTRUMENTS & COMPUTERS, 1992, 24 (02) :190-204