|
|
|
|
LEADER |
00000nam a22000002u 4500 |
001 |
in00004129663 |
003 |
ERIC |
005 |
20220616050153.0 |
007 |
he u||024|||| |
008 |
981104s1998 xx ||| bt ||| | eng d |
035 |
|
|
|a ED427043 Microfiche
|
040 |
|
|
|a ericd
|c ericd
|d MvI
|d UtOrBLW
|
049 |
0 |
0 |
|a EEM#
|
099 |
|
|
|a ED427043 Microfiche
|
100 |
1 |
|
|a Barnette, J. Jackson.
|0 http://id.loc.gov/authorities/names/no91011717
|
245 |
1 |
4 |
|a The Tukey Honestly Significant Difference Procedure and Its Control of the Type I Error-Rate /
|c J. Jackson Barnette and James E. McLean.
|
260 |
|
|
|a [Place of publication not identified] :
|b Distributed by ERIC Clearinghouse,
|c 1998.
|
300 |
|
|
|a 15 pages
|
336 |
|
|
|a text
|b txt
|2 rdacontent
|
337 |
|
|
|a microform
|b h
|2 rdamedia
|
338 |
|
|
|a microfiche
|b he
|2 rdacarrier
|
500 |
|
|
|a ERIC Note: Paper presented at the Annual Meeting of the Mid-South Educational Research Association (27th, New Orleans, LA, November 4-6, 1998).
|5 ericd
|
520 |
|
|
|a Tukey's Honestly Significant Difference (HSD) procedure (J. Tukey, 1953) is probably the most recommended and used procedure for controlling Type I error rate when making multiple pairwise comparisons as follow-ups to a significant omnibus F test. This study compared observed Type I errors with nominal alphas of 0.01, 0.05, and 0.10 compared for various sample sizes and numbers of groups. Monte Carlo methods were used to generate replications expected to provide 0.95 confidence intervals of +/- 0.001 around the nominal alphas of 0.10, 0.05, and 0.01 for 42 combinations of n (5, 10, 15, 20, 30, 60, and 100) and numbers of groups (3, 4, 5, 6, 8, and 10). Means and standard deviations of observed Type I error rates and percentages of observed Type I errors falling below, within, and above the 0.95 confidence intervals were determined for total number of Type I errors. The results indicate that HSD is conservative relative to experimentwise Type I error control across all alpha levels, sample sizes, and number of groups. However, when per-experiment (total Type I errors) is of interest, HSD was liberal at alpha of 0.10 and 0.05, but was very conservative when alpha was 0.01. Results also point out the differences inherent in selection of a Type I error mode of control. Differences between per-experiment and experimentwise Type I error control was mostly a function of the number of groups being compared. As the number of groups increased, the difference between per-experiment and experimentwise error proportions increased. However, sample size was also a significant predictor; as sample size increased, the difference decreased. (Contains 9 tables and 14 references.) (Author/SLD)
|
533 |
|
|
|a Microfiche.
|b [Washington D.C.]:
|c ERIC Clearinghouse
|e microfiches : positive.
|
500 |
|
|
|a Microform.
|
650 |
1 |
7 |
|a Comparative Analysis.
|2 ericd
|
650 |
1 |
7 |
|a Error of Measurement.
|2 ericd
|
650 |
0 |
7 |
|a Monte Carlo Methods.
|2 ericd
|
650 |
1 |
7 |
|a Research Methodology.
|2 ericd
|
650 |
1 |
7 |
|a Sample Size.
|2 ericd
|
650 |
0 |
7 |
|a Simulation.
|2 ericd
|
653 |
1 |
|
|a Tukey Statistic
|a Type I Errors
|
655 |
|
7 |
|a Reports, Research.
|2 ericd
|
655 |
|
7 |
|a Speeches/Meeting Papers.
|2 ericd
|
700 |
1 |
|
|a McLean, James E.,
|e author.
|0 http://id.loc.gov/authorities/names/n84205126
|
907 |
|
|
|y .b64770060
|b 211123
|c 081227
|
998 |
|
|
|a mc
|b 081227
|c m
|d a
|e -
|f eng
|g xx
|h 4
|i 1
|
982 |
|
|
|a no_backstage
|
999 |
f |
f |
|i 9b0b0045-fcd6-55c7-8993-084ba461f616
|s 1ccfdced-d7d2-50c9-924d-e35d38ee9ef7
|t 0
|
952 |
f |
f |
|p Non-Circulating
|a Michigan State University-Library of Michigan
|b Michigan State University
|c MSU Microforms
|d MSU Microforms, 2 West
|t 0
|e ED427043 Microfiche
|h Other scheme
|i Microform (Microfilm/Microfiche)
|n 1
|